Data Brokers & Beyond: Navigating New Jersey’s Data Broker & “Data Collector” Registration Law

Co-authored with Kelly Brandmeyer, FPF U.S. Policy Intern

In a two-day span from June 28 to June 30, the New Jersey legislature introduced and passed A5328, amending New Jersey’s comprehensive privacy law and establishing new data broker and “data collector” registration requirements. The new data broker law has a uniquely broad scope and high financial costs. The law requires not only data brokers to register annually, but also “data collectors”—businesses that have a direct relationship with a consumer but sell personal data to data brokers. The law’s fees and penalties also extend beyond those in existing state data broker laws, with registration fees ranging as high as $1.5M annually and penalties of  $2,500 per day for registration-related violations and $50,000 per record for prohibited sales of sensitive data. 

Governor Sherrill approved the bill on June 30, and the law took effect immediately, except for the creation of the “data broker” and “data collector” registry by the Division of Consumer Affairs in the Department of Law and Public Safety, which will become operative 270 days after enactment. This blog post provides an overview of the bill, including the changes made to the comprehensive privacy law as well as the definitions and requirements of the new data broker registration requirements.

Comprehensive Privacy Law Amendment

This bill amends N.J. Stat. Ann. § C.56:8-166.12 (i.e., controllers’ duties) to prohibit a controller from selling sensitive data. Notably, this prohibition would apply “to all individuals or legal entities regardless of the number of consumers whose data the individual or entity controls or processes.” This means that the law’s applicability thresholds—i.e., limited scope to controllers who control or process the personal data of (1) at least 100K consumers or (2) of at least 25K consumers and who derive revenue from the sale of personal data—do not apply to this requirement, so small- and medium-sized entities who are otherwise outside the scope of the law would be subject to the prohibition. If a controller violates this provision and is either a “data broker” or “data collector” under the new data broker registration provisions (see below), then the civil penalty is $50,000 per record sold, offered for sale, or licensed. (A5328, §§ 1 & 5.)

This is consistent with a legislative trend. Maryland banned the sale of sensitive data when it passed its law in 2024. Since then, Oregon, Virginia, and Connecticut have amended their comprehensive privacy laws to ban the sale of specific types of sensitive data. 

Definitions and Scope: New Data Broker Requirements

The new requirements under this bill are scoped to data brokers and data collectors.

• “Data broker” is defined as a person or entity that “knowingly collects or purchases the personal data of a consumer with whom the person or legal entity does not have a direct relationship and sells or licenses that data to a third party.” The definition includes an illustrative list of direct relationships, including a “past or present . . . customer, client, subscriber, or user of the person or legal entity’s goods or services.” This definition is substantially similar to the definition that Vermont’s law had prior to being amended this year. Oregon’s data broker registration law includes the closest active definition. The definition clarifies that a processor is not a third party if “disclosure of personal data to the processor is solely to process the personal data on the data broker’s or data controller’s behalf.” (A5328, § 2(a).)
• “Data Collector” is defined as “a business, or units of a business, separately or together, that knowingly:  (1) collect the personal data of a consumer with whom the data collector has a direct relationship; and (2) sell or license such personal data to a data broker.” (A5328, § 2(a).)
• Existing definitions under New Jersey’s comprehensive privacy law—including consumer, de-identified data, personal data, precise geolocation data, process, processor, publicly available information, sale/sell, and sensitive data—are reproduced rather than cross-referenced. Although these definitions largely mirror those under the NJDPA, with conforming changes such as changing “controller” to “data broker or data collector,” there are a few other changes. “Sale” is defined as “the sharing, disclosing, or transferring of personal data for monetary or other valuable consideration” and, unlike under the comprehensive privacy law, this definition contains no exclusion list, broadening its scope. For “de-identified data,” the contractual obligations that must be imposed on recipients of de-identified data have an added requirement that a data broker or data collector must “enforce[ ] or otherwise ensure[ ] compliance with” the contractual obligations. (A5328, § 2(a).)

Entity- and Data-Level Exemptions: Different sections of the bill have distinct exemptions. 

Registration Requirements 

Data brokers and data collectors engaged in selling or licensing personal data of New Jersey consumers will be required to annually register with the Division of Consumer Affairs in the Department of Law and Public Safety (“Division”), which will then establish and maintain a public registry of data brokers and data collectors. Similar to existing state data broker registration laws, registrants will have to pay a fee and provide required information with their registration application. 

The registration fees are much higher than under other data broker registration laws, and they increase based on the number of New Jersey consumers whose data are collected and sold or licensed. The lower end of the spectrum is $5,000 for data brokers that sell or license (or data collectors that collect and sell or license to a data broker) the personal data of 100,000 or fewer consumers. The highest fee is $1,500,000 for entities that sell or license the personal data of 4.5 million or more consumers. Critically, the law does not specify a specific deadline or timeframe for data brokers and data collectors to register other than “annually.” The law also does not specify the relevant timeframe for determining the relevant number of consumers whose personal data are collected and sold or licensed for determining the applicable registration fee.

There are nine categories of information that a data broker or data collector must provide with its application. These include information about: the business, such as physical, email, and website addresses; opt-outs offered; consumers’ ability to delete their data; limitations on opt-outs; any credentialing process for purchasers of data; a history of data breaches and cybersecurity events affecting the business; “data collection practices, databases, sales activities, and opt-out methods that are applicable to” data of minors (under 18); anything the Division “deems appropriate” to implement; and processors who process personal data on behalf of the data broker or data collector. Compared to the existing state data broker registration laws, these disclosures are substantial—more detailed and numerous than under some of the laws, though not as intensive as California’s Delete Act. 

Entity- and Data-Level Exemptions: This section of the bill includes a number of exemptions, including for: protected health information under HIPAA; financial institutions, data, and affiliates subject to GLBA; secondary market institutions; certain insurance institutions and insurance-support organizations; personal data sold by the New Jersey Motor Vehicle Commission as permitted by the federal DPPA; personal data collected, processed, sold, or disclosed by a consumer reporting agency if authorized under FCRA; state agencies and political subdivisions (or instrumentalities of political subdivisions); personal data collected, processed, or disclosed as part of research meeting certain safeguards under federal law; and national securities associations registered pursuant to the Section 15A of the Securities Exchange Act. 

The bill also exempts certain activities, such as developing or maintaining a third-party e-commerce or application platform, providing 411 directory assistance or directory information services, or providing publicly available information for certain purposes. However, an entity engaged in those activities is considered a data broker if it sells or licenses data to third parties in a way that is not incidental to one of those exempted activities. There are additional exemptions for (1) nonprofits providing enrollment data reporting services on behalf of postsecondary educational institutions and (2) providing title and settlement services.

Prohibition on Selling Sensitive Data

The bill prohibits data brokers or data collectors from selling or licensing sensitive data “to any other individual or entity.” This is a broader prohibition than under the comprehensive privacy law because the relevant definition of sale lacks common exceptions for disclosure of personal data to a processor, disclosure of personal data to a third party for the purposes of providing a product or service requested by the consumer, the disclosure or transfer of personal data to an affiliate, the intentional disclosure of personal data by a consumer to the general public through a mass media channel, or the disclosure or transfer of personal data to a third party as part of a merger, acquisition, or bankruptcy. (A5328, § 3(a).)

Entity- and Data-Level Exemptions: This section is subject to a separate set of exemptions than the prior section concerning registration. These exemptions appear to be substantially similar to the ones above, apart from the absence of exemptions for certain enumerated activities such as developing or maintaining a third-party e-commerce or application platform, providing 411 directory assistance or directory information services, or providing publicly available information for certain purposes.  (A5328, § 3(b).)

Enforcement

Failure to register—or failure to submit or update required information—will result in a civil penalty of $2,500 per day for a data broker or data collector, in addition to the registration fee. It is notable that these penalties are uncapped. Oregon’s data broker registration law, for example, caps annual fees at $10,000. In contrast, noncompliance with New Jersey’s registration requirement could generate up to $912,500 in fees after 365 days. 

Any data broker or data collector that sells, offers for sale, or licenses sensitive data will be subject to a civil penalty of $50,000 per record sold, offered for sale, or licensed. (A5328, §§ 4 & 5.)

Construction and Rulemaking

The bill provides that the data broker and data collector requirements apply “in addition to and not in lieu of the provisions of” New Jersey’s comprehensive privacy law. Like with the comprehensive privacy law, the Director of the Division of Consumer Affairs has rulemaking authority. (A5328, §§ 6 & 7.)

FPF Hosts Frontiers Workshop on Privacy, AI, and Emerging Infrastructure

On June 10, 2026, the FPF Center for Artificial Intelligence convened a Frontiers Workshop in Washington, DC. Held as part of FPF’s National Science Foundation (NSF) and the Department of Energy (DoE)-funded Privacy-Enhancing Technologies (PETs) Research Coordination Network, the workshop brought together privacy and frontier AI practitioners to examine challenges at the intersection of data governance and AI systems.

Across three sessions, participants explored the technical infrastructure needed to responsibly deploy AI tools in sensitive data environments. Discussions centered on the systems, protocols, and evaluation methods that enable privacy-preserving AI in practice, including: the infrastructure that supports AI deployment; approaches for evaluating AI systems without exposing sensitive information; and the ways AI assistants collect, process, and reveal information about their users.

The workshop featured presentations from Andrew Gruen, CEO at Working Paper and FPF Senior Fellow; Bennett Hillenbrand, President and CPO at Working Paper and Head of Product at MLCommons AIRR; and Libby Hemphill, Associate Professor at the Inter-university Consortium for Political and Social Research (ICPSR), University of Michigan; each highlighting a different aspect of the problem space:

• E-waste as a privacy-enhancing technology: Libby Hemphill (ICPSR) examined the challenges researchers face when using AI tools on sensitive datasets governed by institutional protocols, IRB rules, and regulations such as the Health Insurance Portability and Accountability Act (HIPAA) and the GDPR. Because these requirements often prohibit sending sensitive data to commercial AI services, organizations rely on compliant alternatives, such as cloud and local data centers, which are costly and carbon intensive. Hemphill presented an alternative approach using repurposed smartphones, run as air-gapped local clusters, demonstrating how existing hardware could support private, auditable, portable, and sustainable on-premises AI. 
• Federated evaluation with MedPerf: Bennett Hillenbrand (MLCommons) focused on a common challenge in AI evaluation: balancing the privacy requirements of data holders, the intellectual property interests of model providers, and the integrity of the benchmark itself. Hillenbrand demonstrated how,  with confidential computing, evaluations can be performed where the data resides without exposing either the underlying data or the model itself, returning only test results. 
• Model Context Protocol (MCP) servers and the instrumented document: Andrew Gruen (Working Paper) concluded the presentations with a live demonstration of how AI can turn a static privacy policy into an interactive tool that answers users’ questions. The demonstration also highlighted an important privacy consideration: the same systems that enable interactive assistance can also infer and log unexpected information about users. 

Across all three sessions, the binding constraints involved infrastructure and protocol rather than policy. When AI is brought into sensitive data settings, the decisive controls increasingly live in the physical substrate (what hardware runs the model and where), the execution environment (whether data and tests are exposed during evaluation), and the interface layer (what an AI assistant discloses about its user). As one presenter put it, physical security beats policy: a contract can be circumvented, but physics cannot. The corollary, recurring throughout, is that the same instrumentation that makes these systems governable also generates new and sensitive data — insight and liability arrive together.

The workshop concluded with a collaborative discussion of outstanding concerns for the AI governance community: governance of the “output space” in federated evaluation, standards for local AI inference, user visibility and authorization, the reliability of self-reported intent data in the presence of persona manipulation, and approaches for measuring system reliability that could support market-based governance.

Interested in receiving more information about events like this? Email us at [email protected].

The Research Coordination Network (RCN) for Privacy-Preserving Data Sharing and Analytics is supported by the U.S. National Science Foundation (Award #2413978) and the Department of Energy (Award #DE-SC0024884).

FPF’s 2026 DC Privacy Forum: Leading Voices in AI, Privacy and Emerging Technology

By Paige Garvin, FPF Communications Intern

The Future of Privacy Forum hosted its third annual DC Privacy Forum: Advancing Principled Data Protection, AI, and Digital Governance Practices on June 10th, 2026. This year’s Forum gathered government officials, academics, civil society representatives, and privacy professionals to discuss developments in AI governance, privacy regulation, youth online safety, personalization, AgeTech, and other emerging digital technology issues. 

Through keynote remarks, panel discussions, debates, and lightning talks, participants explored how policymakers, industry leaders, and researchers can address evolving privacy challenges while supporting innovation. 

Government Leadership in Technology Innovation and Privacy Protection

FPF CEO Jules Polonetsky opened the Forum before welcoming FPF Board President Alan Raul to introduce the keynote speaker, Congressman John Joyce, M.D., representative of Pennsylvania’s 13th Congressional District. 

Congressman Joyce reflected on his work advancing federal privacy legislation, including his role as a lead sponsor of the Securing and Establishing Consumer Uniform Rights and Enforcement over Data Act (SECURE Data Act). Although he noted that he entered privacy policymaking without a technical or legal background in the field, he emphasized the value of stakeholder engagement in shaping effective privacy legislation. Over the past 18 months, he has participated in hundreds of meetings and received extensive public feedback on privacy reform proposals. 

“Your participation allows the SECURE Data Act to be the legislation that can ultimately be that successful safeguard, that successful guideline, that successful piece of legislation that America needs,” Congressman Joyce detailed to the Forum attendees. 

He also highlighted the SECURE Data Act’s focus on consumer protections, concise language, and centralized enforcement. He expressed optimism about the prospects for federal privacy legislation and emphasized that meaningful consumer protections and innovation can coexist. 

Concluding his keynote, Congressman Joyce reaffirmed his commitment to advancing privacy legislation and thanked FPF and Forum participants for contributing to the ongoing policy discussion. 

Geostrategy of Regulation: Navigating Digital Sovereignty

The opening panel, “Geostrategy of Regulation: Understanding and Responding to the Global Demand for Digital Data Sovereignty,” examined the growing influence of digital sovereignty initiatives worldwide. Moderated by Gabby Miller, AI and Tech Congressional Reporter at Politico, the discussion featured Pablo Chavez, Adjunct Senior Fellow, Technology and National Security Program, Center for a New American Security (CNAS), Bill Guidera, Deputy Assistant Secretary for Services at The Trade Administration, Cameron Kerry, Ann R. and Andrew H. Tisch Distinguished Fellow at the Brookings Institution and Kenton Thibaut, Senior Resident China Fellow, Atlantic Council, at the Digital Forensic Research Lab (DFRLab).

Panelists explored implications of regulatory fragmentation, the challenges U.S. technology companies face in global markets, and opportunities to strengthen trust in American technology. 

The Future of AgeTech, Privacy, Autonomy, and Aging 

Following a coffee networking break, Jordan Wrigley, FPF Senior Technologist, welcomed the second panel, “The Autonomy Equation: Tech, Privacy and the Future of Aging”. 

Panelists Debra Berlyn, FPF Board Director and Executive Director at Project GOAL, Ian Hartman-O’Connell, Senior Director, Policy Integration at AARP, Dr. Rachele Hendricks-Stirrup, Research Director, Real-World Evidence (RWE) at Duke-Margolis Institute for Health Policy, and Jeff Roby, Director, Enterprise Privacy at Best Buy, discussed the growing role of AgeTech in supporting older adults. The conversation focused on how technologies can promote independence while preserving privacy, autonomy, and user control over the collection of personal information. 

Participants emphasized the importance of designing technologies that are accessible, understandable, and responsive to the needs of older adults. As Berlyn noted, privacy protections must be something that older adults “can touch and see.” 

Enforcement as the Foundation of Digital Rights

FPF Senior Policy Counsel, Jordan Francis, delivered a lightning talk on the role of enforcement in privacy governance. Francis presented the argument that rights and protections are meaningful only when supported by effective implementation and enforcement mechanisms.

He highlighted the continued expansion of privacy legislation across the United States and discussed the importance of ensuring compliance in practice. Francis then introduced the panel “State of U.S. Privacy and AI Regulation,” moderated by Tatiana Rice, FPF Senior Director for U.S. Legislation. 

Rice, joined by Hanna Abrams, Assistant Attorney General at the Office of the Attorney General of Maryland, Charlie Bullock, Senior Research Fellow at the Institute for Law & AI, Senator James Maroney, State Senator, Connecticut General Assembly and Cristy Phillips, Special Counsel, Economic Justice Division at the Office of the New York Attorney General discussed developments in privacy and AI regulation, including data minimization, youth online safety, chatbot governance, and the relationship between state and federal policymaking. Throughout the discussion, speakers emphasized the importance of continued collaboration among policymakers, regulators, technologists, and industry stakeholders. 

The Changing Global Landscape of Privacy and AI 

Following a networking lunch sponsored by Ropes & Gray, Gabriela Zanfir-Fortuna, FPF Vice President of Global Privacy, welcomed attendees back for a discussion on global developments in privacy and AI governance. 

The panel featured Cari Benn, Chief Privacy Officer at Microsoft Corporation, John Miller, Executive Vice President of Policy and General Counsel at the Information Technology Industry Council, Hilary Wandall, Chief Ethics and Compliance Officer at Dun and Bradstreet, and Justin B. Weiss, Senior Counsel and Senior Director at Crowell & Moring LLP. Panelists examined differences in how privacy and AI regulation are approached globally and considered lessons U.S. policymakers can draw from international frameworks. 

The discussion highlighted developments in the United Kingdom, South Korea, Singapore, and Latin America, with speakers emphasizing that effective governance requires not only legal and technical expertise but also an understanding of cultural and historical contexts. Panelists agreed that global privacy and AI governance will continue to evolve as governments seek to balance innovation, economic competitiveness, and individual rights. 

The Big Debate: Chatbots and Youth

Daniel Hales, FPF Policy Counsel for U.S. Legislation, brought energy to the second half of the Forum with FPF’s recurring debate series, an interactive format that allows attendees to participate via online voting before, during, and after hearing competing perspectives. 

This year’s debate focused on youth engagement with chatbots. Meg Leta Jones, Provost’s Distinguished Associate Professor and Cartoonist at Georgetown University, argued in favor of regulations and restrictions. She cited concerns about manipulation, social isolation, sexual exploitation, and mental health risks, arguing that policymakers should establish guardrails to protect children and support parents. 

Jennifer Huddleston, Senior Fellow, Technology Policy at the Cato Institute, argued against additional regulation, emphasizing parental choice, privacy considerations, and the challenges of implementing age-verification requirements. She suggested that many proposed legal solutions may create new privacy concerns while failing to address underlying issues effectively. 

Although audience members initially and finally voted in favor of increased regulation, the debate’s outcome was determined by the percentage change in opinion throughout the session. Huddleston – who won a debate at FPF’s 2024 DC Privacy Forum: AI Forward – ultimately persuaded a larger share of attendees to move in her direction and was once again awarded the FPF GOAT Trophy.

Understanding Adult Perspectives Toward Age Verification 

Building on themes discussed during the debate, Carnegie Mellon University Professor and CyLab Security and Privacy Institute Director, Lorrie Cranor, presented findings from research titled “User (Non-)Compliance with Age Verification: Evidence from a Deceptive Web Experiment”. 

Cranor noted that while substantial research has focused on age verification for minors, less attention has been paid to how adults respond to different verification methods. Her study explored how factors such as verification techniques, data handling practices, and privacy concerns influence users’ willingness to access age-restricted content online. 

Drawing on responses from approximately 1,600 participants, the research found that willingness to engage decreased as verification methods became more invasive. While nearly all participants were willing to confirm their age through simple checkboxes, fewer than one-quarter indicated they would upload a government-issued identification document. The findings highlighted the importance of considering privacy, usability, and trust when designing age verification systems. 

AI and The Workforce

Stacey Gray, FPF Senior Director of Artificial Intelligence, introduced Taylor Stockton, Chief Innovation Officer at the Department of Labor, for a fireside chat on AI adoption in the workplace. 

Stockton discussed the increasing use of AI across federal agencies and the opportunities these technologies present for improving government services. He emphasized the importance of cybersecurity, workforce training, and risk-based governance frameworks as organizations integrate AI into existing operations.  

Stockton also addressed public perceptions of AI, encouraging a shift from viewing AI primarily through the lens of risk toward a more balanced understanding that recognizes both challenges and opportunities. He noted that adaptation and workforce development will be critical as technological change continues to accelerate. 

The discussion was followed by a panel featuring Aditya Bharadwaj, Senior Director, Assistant General Counsel at UKG, Barbara Cosgrove, Vice President and Chief Privacy Officer at Workday, Sara Harrington, Vice President, Legal (Data, AI and Privacy) at LinkedIn, and Sheila Jambekar, Senior Vice President, Chief Privacy Officer at DayForce. Panelists examined the use of AI in hiring and workforce management, discussing transparency, accountability, human oversight, and responsible deployment practices. 

Following the panel, lightning talks from Jameson Spivack, FPF Deputy Director for Artificial Intelligence, and Stacey Gray, addressed emerging areas of focus in AI. Spivack discussed emerging questions surrounding data-driven pricing practices, while Gray explored topics of spatial intelligence, bystander privacy, and governance considerations for autonomous AI agents.

Personalization and Youth Online 

The concluding panel of the day focused on personalization and youth online. The panel started off with the debut of the new FPF report, Personalization and Youth Online: Assessing Benefits, Risks, and Safeguards, co-authored by FPF’s Daniel Hales and Holly Hawkins, which provides an overview of current policy discussions surrounding personalization of digital experiences. 

Hales moderated the panel and noted that conversations about personalization should focus not only on what information organizations collect, but also on how that information is used and who is affected by those uses. He observed that policymakers continue to evaluate potential safeguards while balancing the benefits that personalization can provide.  

Joining the discussion were Holly Hawkins, FPF Director of Youth Policy, Jared Bomberg, U.S. Policy Lead, Privacy and Data Strategy at Google, Dona J. Fraser, Senior Vice President, Privacy Initiatives at BBB National Programs, and David Lieber, Head of Privacy, Public Policy for the Americas at TikTok. 

Panelists explored both the benefits and risks associated with personalization technologies. Speakers highlighted how personalization can help users discover relevant content, improve accessibility, and create age-appropriate experiences. At the same time, they discussed concerns related to excessive data collection, profiling, autonomy, and transparency. 

While panelists approached these issues from different perspectives, there was general agreement that personalization presents both opportunities and challenges that require thoughtful, context-specific policy solutions.

Looking Ahead 

FPF Vice President for U.S. Policy Matthew Reisman delivered closing remarks, thanking attendees, speakers, sponsors, and partners for contributing to the Forum’s success. 

Throughout the day, participants examined a wide range of issues shaping the future of privacy, artificial intelligence, and digital governance. Discussions highlighted the importance of collaboration among policymakers, industry leaders, academics, civil society organizations, and technical experts in addressing emerging challenges and opportunities. 

Thank you to all who participated in our annual DC Privacy Forum: Advancing Principled Data Protection, AI, and Digital Governance Practices! This year’s DC Privacy Forum was made possible thanks to our sponsors FTI Consulting, SafePorter, Greenberg & Traurig, and Ropes & Gray.

We hope to see you next year! For updates on FPF work, please visit FPF.org for all our reports, publications, and infographics. Follow us on LinkedIn, Instagram, X, and YouTube, and subscribe to our newsletter for the latest.

Understanding Data Embassies and Corridors

The following is a guest post to the FPF blog authored by Yeong Zee Kin, Chief Executive of the Singapore Academy of Law and FPF Senior Fellow. The guest post reflects the opinion of the author only and does not necessarily reflect the position or views of FPF and our stakeholder communities. FPF provides this platform to foster diverse perspectives and informed discussion.

Over the past few years, geopolitical contestations have increased the rhetoric over data sovereignty. Data sovereignty views data as another dimension of the state’s sovereignty that needs to be safeguarded from exploitation against the interests of the state. One natural response is to mandate the localization of data. 

The intensification of geopolitical tensions has also changed the tone of trade discussions. Countries that were once strident advocates of free trade have, to varying degrees, introduced some form of data localization policies. One example of recent regulatory developments that limit transfers of personal data to specific countries is from the US. The White House Executive Order 14117 has now been implemented into final rules by the Department of Justice¹; and the US Congress has also passed the Protecting Americans’ Data from Foreign Adversaries Act of 2024². These laws prohibit the transfer of personal data to certain countries that are deemed to be adversaries of the US in the interest of safeguarding US national security. At the same time, the US remains a strong advocate for cross-border transfer mechanisms such as the Global Cross Border Privacy Rules (CBPR) and Privacy Recognition for Processors (PRP). 

With the increase in the number of data localization measures globally³, the concept of data embassy has been put forward as a solution by both governments and businesses. But are data embassies an appropriate solution? This paper examines the origin and varieties of data embassies, discusses the concept’s limitations to address data localization challenges, and proffers an alternative modelled after special economic zones.

Data localization 

Data localization is no longer associated only with trade protectionism but has taken on a security complexion. Data localization policies can be mapped on two dimensions — they may either prohibit the export of data, mandate local storage and processing of data, or both. It is possible to construct a two-dimensional matrix. 

The earliest implementations of data localization regulations were largely motivated by economic policies. For example, the belief that requirements for local storage and processing of data will spur investments in data centres and digital communications infrastructure; or the creation of data analytics and processing jobs that allow the domestic workforce to upskill and help develop its digital economy. However, research shows that data localization may have the countervailing effect of increasing data management costs by 15 – 55%⁴, thereby subtracting from the perceived industry development benefits by decreasing trade output, increasing costs for downstream industries, and decreasing productivity⁵.

Another reason that has been given for requiring local storage of data is to enable (easier) access by law enforcement and judicial authorities. In more recent years, security concerns have also given rise to localization policies that prohibit the export of data that can be potentially exploited to effect socio-economic harm or enable attacks on critical information infrastructure or foundational digital infrastructure. 

Efforts to preserve the free flow of data may be found in trade agreements, particularly digital trade agreements, and global norm-setting initiatives such as the G7’s efforts to ensure data free flow with trust. In gist, these policies seek to reduce hurdles to cross-border data transfers as indirect trade barriers by prohibiting signatories from imposing requirements of local processing or storage of data as a condition of doing business. Within these frameworks, restrictions to data transfers may only be imposed if they are necessary for achieving legitimate public policy objectives and even then, their scope must be proportionate to the identified harm.

Data embassy as a potential solution to data localization

Private international law supports the choice of law that governs private rights between commercial entities, the form of dispute resolution (e.g. litigation or arbitration), and choice of forum (and hence, the governing procedural rules). However, data localization requirements are in the realm of public law that cannot be contracted out of. The data embassy was initially developed as a government-to-government (G2G) arrangement. It has since seized the attention of businesses as a solution to circumvent data localization requirements. The primary motivation is to extend domestic laws and standards of protection to data that has been exported. What if data that has been exported is somehow still subject to the laws of the jurisdiction they originated from? If domestic laws and standards of protection follow the exported data, these overseas repositories of data will be like embassies in foreign countries that, despite their location overseas, are still treated as part of the home jurisdiction. 

This is attractive to businesses for a number of reasons. In one scenario, data centres or data processors hoping to win contracts may wish to assure their overseas customers that their data will be managed according to their laws. This typically plays out in scenarios where the data centre or processor is located in a lower cost jurisdiction that is perceived to have lower standards of data protection. From the perspective of the host state (i.e., where the data centre is built), data embassies have industry development potential by attracting foreign direct investments for the development of communications and digital infrastructure. In another scenario, corporations in a country with data localization requirements seeking to conduct business overseas may proffer the data embassy as a way of assuring domestic regulators that exported data will remain subject to their regulatory requirements and within their reach, while concomitantly inaccessible by the government of the host state. From the exporting state’s perspective, this solution could work if it is assured of cooperation from the host state when access is required; while from the host state’s perspective, it must be prepared to permit intrusion into its sovereignty. 

The origins of the concept and limitations of data embassies⁶

The concept of “data embassy”, while intellectually appealing, is based on a flawed understanding of how embassies function. A misapprehension that belies the issues associated with data embassies as a solution to data localization policies. Contrary to common misconception, embassies are not pockets of foreign sovereignty. While the sending state may own the land upon which the embassy stands, it is nevertheless subject to the laws of the receiving state. 

By international law and custom documented in the Vienna Convention on Diplomatic Relations (1961), the receiving state extends certain privileges to the embassy. First, diplomatic premises are inviolable in that the receiving state will refrain from exercising powers of search unless there is consent. Concomitantly, the receiving state has positive duties to protect the mission⁷. Second, the receiving state likewise refrains from exercising criminal jurisdiction or powers of arrest and detention over diplomatic agents⁸. Third, official communications from the mission are also protected⁹. Subject to these privileges, the law of the land applies. The mission is liable for injuries and harms that befall guests; and administrative, technical and service staff of the mission may not undertake unlawful activities within the embassy with impunity. Diplomatic agents who commit crimes will face prosecution in their home country even while they are not subject to criminal jurisdiction in the receiving country. 

Hence, the assumption that data stored in a data embassy is governed by the laws of the sending state (i.e., the state where the data originates) is deeply flawed. If we are to faithfully apply what we may legitimately glean from the way real-world embassies operate to data hosted in a data embassy, we may only arrive at the position that the receiving state (i.e., the state where the data has been exported to) accepts that it cannot access the data, must protect data-at-rest and ensure the security of data-in-transit. The quid pro quo is that the sending state must investigate data incidents and take appropriate enforcement action – not something that can be easily accomplished remotely. To be clear, these are public law obligations and not obligations that can be negotiated in the realm of private law.

Presently, there are two primary data embassy models: the security model and the developmental model. While they share a common name, they are very different creatures. 

Data embassy as a concept started with the Estonian implementation of the security model, which is perhaps the closest implementation to a real-world embassy. As a neighbor to a much larger and aggressive country, Estonia had its fair share of experiences with cyberattacks that have been linked to state-sponsored groups. After a particularly invidious incident¹⁰, it struck an arrangement with Luxembourg to host a copy of its public sector data in the latter. As part of the arrangement, Luxembourg undertook to preserve the inviolability of this data center from searches and entry, its protection from intrusion and the confidentiality of communications¹¹. In addition to Estonia, Luxembourg has also entered into a similar arrangement with Monaco¹². These are, in essence, G2G arrangements for off-site backup of government data. They are not particularly helpful for solving the commercial and regulatory compliance concerns of businesses.

The developmental model of data embassies attempts to extend the data embassy concept to attract foreign direct investments. Bahrain’s implementation exemplifies this model. It aims to attract investments in cloud infrastructure and support the development of cloud services. Bahrain passed Decree 56 of 2018 that allows the disapplication of domestic law to content hosted in designated data centers while concomitantly designating the foreign law, competent courts and public authorities that will have exclusive jurisdiction instead¹³. Such a designated data center (i.e., the data embassy) is designed to allow cloud service providers operating there to choose the law that will be applied to customer data stored on their cloud platforms. However, there are a number of challenges with this approach. 

While private international law allows contracting parties freedom of choice over governing law, dispute resolution mechanism and forum, data incidents have both private and public law consequences. Thus, the aftermath of a cybersecurity incident or data breach follows two tracks. First, breaches of contractual data protection and cybersecurity obligations between data center and customer can be enforced through private law. Second, the cybersecurity incident or data breach will also be investigated and enforced by the relevant data protection authority, and cybersecurity and law enforcement agencies of the jurisdiction where this incident occurred. 

It is troubling to conceive of a situation where the receiving state has disapplied its public laws and declines to investigate and take enforcement action. While at the same time, the foreign state whose cybersecurity and data protection laws were chosen to apply does not enforce them because it is ignorant of this choice, has not agreed to take on this role, or cannot practically do so. Offshore enforcement requires the foreign state to extend its investigatory and enforcement powers into the receiving state, which raises additional issues of sovereignty. It also requires the relevant data protection authorities, and cybersecurity and law enforcement agencies of the foreign state to have the capability of conducting investigations and collecting evidence remotely since the data repository is situated in another country. 

This data embassy design also fails to solve data localization issues. The localization requirements operate on the customers of the data center in the country where they are situated. Until and unless the state exempts these customers from localization requirements, they are not able to circumvent such requirements by selecting an overseas data centre that allows them to select their choice of law, even if they choose their own laws, both private and public. Unilateral action by the destination state for exported data does not provide a complete solution.

In an April 2025 public consultation, Saudi Arabia put forward another variation of the developmental model for data embassies under its draft Global AI Hub Law that seeks to support the Kingdom’s ambitions to develop into a global AI hub¹⁴. The building block appears to be a security-styled data embassy for public sector data (referred to as a private hub in the draft law). This may be extended to allow a foreign third-party operator to offer its services to other commercial customers (referred to as an extended hub in the draft law). The third variation allows service providers based in Saudi Arabia to offer hosting services to commercial customers in foreign states under the laws of those foreign states (referred to as a virtual hub in the draft law). G2G agreements are contemplated for both private and extended hubs. It appears that service providers are also required to enter into agreements with the competent authority (for extended hubs) or obtain ex ante ministerial approval (for virtual hubs). It remains to be seen whether the regulatory lacuna that had been discussed in relation to the Bahraini implementation are addressed when the Global AI Hub Law is finalized.

Corridors of trust to facilitate bidirectional data flows¹⁵

If data embassies are intended to provide a solution to data localization requirements, the solution must be bidirectional. It must be emphasized that private international law already permits data centers and customers to select their choice of law and jurisdiction to govern their contractual relationship. But data localization requirements are a matter of public law. Hence, a public law solution is required. The public law solution can take a leaf from arrangements that exist for special economic zones. 

For expediency, let us discard analogies with embassies. This public law solution must recognize that bidirectional data flow is a key design consideration and business requirement. (If all that is required is a secondary site for storing data with occasional repatriation, then perhaps the security model for data embassies is well-suited.) For data centers and their customers, data not only needs to flow between them inter se; data also needs to be transmitted to the end customers of the data center’s customers. For example, an e-commerce marketplace hosts its platform with a cloud service provider, but data also must flow from the marketplace to its end users. 

The public law solution needs to check a number of boxes. First, it needs to support a choice of law. Next, it needs to enable access by data protection authorities, cybersecurity and law enforcement agencies to support investigations and enforcement. Additionally, it should also clarify the rights of access by data subjects and data owners. The solution should be capable of functioning as a free-standing data transfer mechanism that can be deployed to support cross-border trade, such as between special economic zones. In such cases, it lowers compliance costs and promotes trade. Additionally, it can also support limited exemptions to data localization requirements in one or both of the participating states. In this case, it also supports trade by removing non-tariff barriers in the form of data localization requirements.

The first point to be clear about is that although the solution is a public law one, it can operate hand-in-glove with private law solutions. The participating states – figuratively, the two terminal points of this data corridor – must first calibrate their applicable laws and regulations. For example, the corridor can operate between two special economic zones. A feature of the special economic zones is that some laws, particularly those relating to customs and tariffs, are specially designed to promote trade. In like manner, the relevant laws for calibration to facilitate bidirectional data flows in a data corridor are likely to be cybersecurity and data protection laws (including any data localization requirements). One way of achieving this is to reference a neutral international data protection standard for both participating states to calibrate their data protection laws to conform with this standard. This is preferable to bilateral mapping as referencing an independent standard makes it easier to scale, such as when other states seek to join this corridor. It also avoids any uncomfortable qualitative assessments when differences are identified when two laws are compared directly.

If special rules are required after benchmarking against that international standard, they may be especially enacted and limited in application to the participating special economic zones. For example, special rules to recognize industry certifications for data protection or cybersecurity (e.g., ISO 27000 series) or cross-border transfers (e.g., Global CBPR and/or PRP) as meeting the requisite regulatory standards and requirements. In this context, special rules that soften data localization requirements may also be possible (e.g., permitting data export if the designated technical standards are met). Once this is achieved, the data center and its customers may then choose which of the participating state’s laws to apply.

Take the Johore-Singapore special economic zone as an example. Both Malaysia and Singapore are part of ASEAN. ASEAN member states have adopted a set of data protection principles – the ASEAN data protection framework – that can serve as the neutral mapping standard. In addition to the principles, ASEAN has also adopted a digital data governance framework and an AI governance framework. These provide a rich source of standards and practices that support implementation. For cybersecurity standards, there are ample technical industry standards that can be referenced. For cross-border transfers, ASEAN has also endorsed the Global CBPR and PRP certification standard.

The choice of law will govern both the private and public dimensions of the commercial relationship between a data center and a customer. In the event of a private dispute, private international law principles will be applied to respect the choice of law and jurisdiction in the resolution of the dispute by the court of the chosen jurisdiction. Should there be a cybersecurity incident or data breach, there needs to be an effective enforcement cooperation agreement between the two participating states. An enforcement cooperation agreement will deal with issues such as a protocol for mutual assistance in acquiring evidence and witness statements. Truth be told, the choice of law is limited to the participating states.

Picking a law of a different state will not work for the public law dimension of this solution for obvious reasons. (To be clear, the data centre’s customers and their end users are not thus restricted.) 

Access to data by regulatory, law enforcement and judicial authorities is another area of concern. There are multiple stakeholders with different interests. The state wants access when it needs to, in order to enforce its laws effectively. Data subjects and data owners want the assurance that access by the government is lawful and subject to independent oversight. Data center operators and cloud service providers want clarity of their roles and responsibilities so that they are not laden with unreasonable or numerous requests. In this space, there are also international and industry standards that can provide an independent and neutral standard that participating states in the data corridor may use as a mapping standard¹⁶. The OECD declaration on government access to data is an example of an international standard, whilst the Trusted Cloud Principles by the Trusted Cloud Initiative is an example of an industry standard¹⁷.

Conclusion

As data emerges as a pivotal factor of production in the 4th industrial revolution—mirroring the regulatory trajectories once charted for land, labor, and capital—it is inevitable that regulatory frameworks around data will intensify. This paper has explored how increasing data localization requirements, fuelled by shifting geopolitical landscapes and heightened security concerns, present significant challenges to the seamless flow of information essential for the digital economy. In response, policy innovations such as data embassies and data corridors offer promising, albeit nascent, pathways to reconcile the imperatives of cross-border data transfers with legitimate governmental interests. These concepts demand rigorous debate, targeted pilot initiatives, and continual refinement to ensure they effectively address both commercial needs and regulatory oversight. Ultimately, striking a careful balance between enabling global data flows and safeguarding national interests will be crucial to harnessing the full potential of the digital economy in this new era.

Yeong Zee Kin¹⁸

1. 28 C.F.R. Part 202. ↩︎
2. 15 U.S. Code Chapter 123. ↩︎
3. The nature, evolution and potential implications of data localisation measures (10 November 2023) OECD, pp 12 – 15. ↩︎
4. The nature, evolution and potential implications of data localisation measures (10 November 2023) OECD, p 3. ↩︎
5. Nigel Cory & Luke Dascoli, “How Barriers to Cross-Border Data Flows Are Spreading Globally, What They Cost, and How to Address Them” (19 July 2021) Information Technology & Innovation Foundation, available at https://itif.org/publications/2021/07/19/how-barriers-cross-border-data-flows-are-spreading-globally-what-they-cost. ↩︎
6. For further reading, see Data Embassies Issues Paper (January 2024) and Data Embassies: Purposes, Features and Limitations (February 2024), Asian Business Law Institute, available at https://abli.asia/abli-publications/abli-data-embassy-issues-paper/ and https://abli.asia/abli-publications/data-embassies-purposes-features-limitations. ↩︎
7. Vienna Convention on Diplomatic Relations (1961), Art 22. ↩︎
8. Vienna Convention on Diplomatic Relations (1961), Arts 29 & 31. ↩︎
9. Vienna Convention on Diplomatic Relations (1961), Art 27.
   ↩︎
10. Emma Savouroux, “A World First: Estonia Opens a ‘Data Embassy’ in Luxembourg” (25 July 2025), available at https://www.blue-europe.eu/analysis-en/short-analysis/a-world-first-estonia-opens-a-data-embassy-in-luxembourg/. ↩︎
11. Agreement between the Republic of Estonia and the Grand Duchy of Luxembourg on the hosting of data and information systems, available at https://www.riigiteataja.ee/aktilisa/2280/3201/8002/Lux_Info_Agreement.pdf.
    ↩︎
12. E-embassies in Luxembourg available at https://luxembourg.public.lu/en/invest/innovation/e-embassies-in-luxembourg.html. ↩︎
13. Legislative Decree No. 56 of 2018 in respect of Providing Cloud Computing Services to Foreign Parties, available at https://www.lloc.gov.bh/FullEn/L5618.docx.
    ↩︎
14. Brian Meenagh, Ksenia Koroleva, and Faisal Imam, Saudi Arabia Pioneers Data Embassies With Publication of Draft Global AI Hub Law, Global Privacy & Security Compliance Blog (18 April 2025), available at https://www.globalprivacyblog.com/2025/04/saudi-arabia-pioneers-data-embassies-with-publication-of-draft-global-ai-hub-law/. 
    ↩︎
15. For a specific design of such a corridor of trust, see https://abli.asia/abli-publications/principles-of-asean-framework-on-crossb-border-cloud-computing/; see also, “ASEAN endorses Malaysia-led Regional Framework on Cross-Border Cloud Computing” (26 February 2026) MDEC https://www.mdec.my/media-release/news-press-release/415/asean-endorses-malaysia-led-regional-framework-on-cross%02border-cloud-computing.
    ↩︎
16. Declaration on Government Access to Personal Data held by Private Sector Entities (14 December 2022) OECD/Legal/0487.
    ↩︎
17.  https://trustedcloudprinciples.com/.
    ↩︎
18. I wish to thank Ms. Catherine Shen for her assistance in reviewing an earlier draft of this paper. ↩︎

Perseverance Pays Off for Vermont Privacy Efforts

Vermont has become the 23rd U.S. state to enact a comprehensive consumer privacy law after Governor Scott signed S.71, the Vermont Data Privacy and Online Surveillance Act (VDPOSA), on June 16. This new law is amongst the broadest in the country, closely resembling the 2025 version of the Connecticut Data Privacy Act (CTDPA). For example, the VDPOSA includes low applicability thresholds, a broad definition of sensitive data, heightened protections for consumer health data, consumer rights to know third parties to whom your personal data is sold and to contest certain profiling decisions, and impact assessments for certain uses of profiling. The law will take effect on January 1, 2028 and be enforced exclusively by the attorney general.

In addition to enacting the VDPOSA, Vermont also passed bills updating the state’s data broker registry (H.211), establishing a direct-to-consumer genetic testing law (H.639), and recognizing a right to neural privacy (H.814). This blog post provides background on Vermont’s privacy legislative efforts in recent years, then covers the law’s scope and key definitions, consumer rights, business obligations, and enforcement provisions. The blog post concludes with a brief overview of other privacy legislation enacted in Vermont this year.

Background

Privacy has been a long time coming in the Green Mountain State. Two years ago, Governor Scott became the first governor to veto a comprehensive consumer privacy bill. That bill, H.121, was an omnibus privacy bill with comprehensive protections and an age-appropriate design code. Had that bill been enacted, the comprehensive privacy provisions would have been amongst some of the broadest and most stringent in the country. In particular, the bill included Maryland-style substantive data minimization requirements, a ban on selling sensitive data, and a limited private right of action (PRA). The legislature tried, but failed, to overturn the veto. 

The legislature continued working on privacy issues in the intervening years. Last year, they enacted the Vermont Age-Appropriate Design Code Act. This year, they finally reached consensus on a comprehensive consumer privacy law as well as an update to the state’s data broker registry, regulation of direct-to-consumer genetic testing companies, and a “right” to “mental and neural data privacy.” Although the law enacted this year diverges from the 2024 effort in notable ways, this law nevertheless incorporates many elements from the broadest and most privacy protective iterations of the Washington Privacy Act framework in the country. 

Scope and Key Definitions

Covered Entities: The law applies to persons who conduct business in Vermont or produce a product or service targeted to Vermont residents and, excluding payment transaction data, annually either (1) control or process the personal data of at least 35,000 consumers, (2) control or process the sensitive data of at least 3,000 consumers, or (3) offer for sale the personal data of at least 3,000 consumers. These thresholds are low compared to those in other states, and it is uncommon to include a threshold tied to processing sensitive data. Like Connecticut’s and Maryland’s laws, the VDPOSA has requirements for consumer health data and consumer health data controllers that are not subject to the same applicability thresholds, instead applying broadly to “a person that conducts business in [Vermont] or a person that produces products or services that are targeted to residents of [Vermont.” This law also addresses any potential conflicts with the Vermont Age-Appropriate Design Code Act (AADCA), providing that the most protective law should control in any situation where that law conflicts with the requirements of this law. (Section 1, § 2415b.)

Definitions: The law’s definitions are generally consistent with the Connecticut model, including aspects of Connecticut’s 2023, 2025, and 2026 amendments. Two definitions worth noting: 

• “Publicly available information” is defined narrowly, excluding information collated and combined to create a consumer profile that is made available via a publicly available website and inferences generated from such profiles. For the consumer rights to access, correction, and portability (see below), publicly available information does not include “information made available for sale,” meaning such data is subject to those rights. This narrow definition of publicly available information—which conversely expands the scope of personal data—is similar to Connecticut’s updated definition from this year’s CTDPA amendments. 
• “Sensitive data” is defined broadly, including less common categories such as data revealing status as nonbinary or transgender, consumer health data, neural data, information derived from genetic or biometric data, certain financial account numbers or login information, or government-issued identification numbers. (Section 1, § 2415a.)

Entity and Data-Level Exemptions: The law includes many of the common entity-level exemptions, including for: certain government entities acting “in the ordinary course of its operation”; a covered entity or business associate under HIPAA (although a “hybrid entity” is not fully subject to the exemption); state or federally chartered banks or credit unions or affiliates or subsidiaries principally engaged in financial activities; certain health care providers and health care facilities under Vermont law; nonprofits established to detect and prevent insurance fraud; and more. Continuing a trend in recent years, the VDPOSA opts for more targeted entity-level exemptions for specific types of financial entities and nonprofits rather than broader exemptions for all GLBA-regulated entities and all nonprofits. 

The law also includes many of the common data-level exemptions, including for: certain health records, patient identifying information, and research data; activities using information for the purpose of evaluating creditworthiness, credit standing, credit capacity, character, general reputation, personal characteristics, or mode of living if “done strictly in accordance with” the FCRA by a consumer reporting agency, furnisher, or person using a consumer report; information collected, processed, or disclosed in accordance with the DPPA or FERPA; data subject to GLBA; protected health information under HIPAA; personal data of a victim or witness of certain crimes (e.g., child abuse, human trafficking) maintained by a victim services organization; and more. (Section 1, § 2415c.)

Exceptions for Common Business Activities: The law includes many exceptions which are consistent with existing state comprehensive privacy laws, including: compliance with federal, state, or municipal laws or regulations; compliance with investigations, subpoenas, or summons; compliance with law enforcement agencies; preventing or detecting security incidents, fraud, or illegal activity; engaging in public or peer-reviewed scientific or statistical research in the public interest that meets required safeguards; internal use of data for product improvement or for internal operations reasonably aligned with the expectations of the consumer; and more. (Section 1, § 2415i.)

Consumer Rights

Consumers have the standard rights to confirm whether a controller is processing their personal data and access that data, correct inaccuracies in their personal data, delete their personal data, obtain a copy of their personal data in a portable format (if technically feasible), and to opt-out of the processing of their personal data for targeted advertising, the sale of personal data, or profiling in furtherance of a decision that produces a legal or similarly significant effect concerning the consumer. These rights contain a few unique or uncommon provisions:

• The right to access extends to inferences about the consumer derived from personal data and the right to know whether a controller or processor is processing the consumer’s personal data for profiling to make a decision that produces legal or similarly significant effects; 
• Consistent with states like Minnesota and Montana, a controller is prohibited from disclosing certain categories of personal data in response to a consumer access request. These include social security numbers, government-issued ID numbers, financial account numbers, health or medical identification numbers, account passwords, security questions or answers thereto, or biometric data. If a controller has collected any of these categories of personal data, it must inform the consumer “with sufficient particularity” that it has done so. 
• If a consumer’s personal data was processed for profiling in furtherance of an automated decision that produced any legal or similarly significant effect, the consumer has the right to, “if feasible,”  question the profiling result, be informed of the reason the profiling resulted in the decision, and review the personal data processed for the profiling. If the decision concerned housing, the consumer can also correct any inaccurate personal data that was processed and have the decision reevaluated based on the corrected data. This right to contest adverse profiling decisions matches that in Connecticut’s law, which was added last year and modeled on Minnesota’s broader right. 
• Consumers have an additional right to obtain a list of the third parties to whom their personal data was sold, or, if the controller does not maintain such a consumer-specific list, then a list of all third parties to whom personal data is sold. 

This law allows consumers to designate an authorized agent to opt out of processing on the consumer’s behalf (including for profiling) and to use an opt-out preference signal to opt out of the sale of personal data or targeted advertising. (Section 1, § 2415d.) 

Business Obligations

Controllers and processors have enumerated responsibilities under the law, including transparency, data minimization, data security, oversight of processors, antidiscrimination, heightened protections for minors, and conducting both data protection assessments and impact assessments. 

Transparency: Controllers must provide consumers with a “reasonably accessible, clear, and meaningful” privacy notice that includes required information under the law, such as categories of data processed, processing purposes, how to exercise rights and appeal decisions, the categories of personal data sold to third parties, and the categories of third parties to whom personal data is sold. (Section 1, § 2415e(c).)

Data Minimization: The law includes procedural data minimization requirements. A controller must: 

• limit the collection of personal data to what is “reasonably necessary and proportionate” in relation to the purposes for which the data are processed, as disclosed to the consumer; 
• obtain consent to process personal data for any “material new” purpose that is “neither reasonably necessary to, nor compatible with,” the disclosed purposes for which the personal data is processed; and 
• not process a consumer’s sensitive data without the consumer’s consent and “unless the processing is reasonably necessary in relation to the purposes for which the sensitive data are collected.” 

Although these provisions are not “substantive data minimization” requirements in the same way that Maryland’s or California’s are, they are slightly unusual. In particular, the “necessary and proportionate” language is a departure from the usual “adequate, relevant, and reasonably necessary” language used in most state laws based on the WPA framework. Only Connecticut uses this same language, and those requirements were added in last year’s CTDPA amendments. Nevertheless, this is still a procedural requirement that ties data collection to the purposes disclosed to the consumer. Also similar to Connecticut, this law explicitly states that a controller cannot sell a consumer’s sensitive data without consent. (Section 1, § 2415e(a).)

Data Security: Controllers are required to “establish, implement, and maintain reasonable administrative, technical, and physical data security practices to protect the confidentiality, integrity, and accessibility of personal data.” (Section 1, § 2415e(a)(2).)

Processors: Controllers must engage in oversight of processors by entering into a contract that meets statutory criteria (e.g., providing instructions for processing data, describing the nature and purpose of the processing, imposing confidentiality). (Section 1, § 2415f.)

Antidiscrimination: Controllers are prohibited from processing personal data in violation of a federal or state law that prohibits unlawful discrimination against consumers. Similar to Connecticut’s 2025 amendment, this law further provides that, for state laws only, any evidence (or lack thereof) of proactive anti-bias testing or similar efforts to avoid processing data in violation of any anti-discrimination law will be relevant to any claim for a violation of such a state law. The law also includes a narrow exception for internal data use in profiling to correct bias. (Section 1, § 2415e(a)(5).)

Consumer Health Protections: Similar to Connecticut’s and Maryland’s law, this law includes heightened protections for consumer health data, such as confidentiality requirements for employee access to consumer health data, a prohibition on geofencing health care facilities for certain purposes (within 1,850 feet), and a prohibition on selling consumer health data without the consumer’s consent. Consumer health data is defined broadly as “any personal data that a controller uses to identify a consumer’s physical or mental health condition, diagnosis, or status,” and it includes reproductive or sexual health data and gender-affirming health data. These protections apply more broadly than the rest of the law to “persons,” notwithstanding the law’s other applicability thresholds. (Section 1, § 2415k.)

Assessments: Like most comprehensive privacy laws, this law requires controllers to conduct and document a data protection assessment for certain processing activities that present a heightened risk of harm to consumers, including: processing personal data for targeted advertising; selling personal data; processing personal data for profiling that presents a reasonably foreseeable risk of substantial injury to consumers, or processing sensitive data. Once again taking inspiration from Connecticut’s 2025 amendment, this law will additionally require a controller to conduct an impact assessment for any profiling conducted for making a decision that produces legal or similarly significant effect. These impact assessments must include information such as: the purpose, intended use, and deployment context of the profiling; analysis on whether the profiling presents a reasonably foreseeable risk of harm; descriptions of inputs and outputs; post-deployment monitoring and user safeguards; and more. The Vermont Attorney General (AG) may request a completed data protection or impact assessment as part of an investigation. (Section 1, § 2415g.)

Enforcement

The VDPOSA will be enforced exclusively by the attorney general. The law includes a permissive cure period of 60 days, allowing the attorney general to issue a cure notice to an alleged violator if the attorney general “determines that a cure is possible.” This cure period will expire on June 30, 2029. Although this law does not include a private right of action (PRA), the legislature added a statement of intent declaring that the attorney general will bear the burden of enforcing the law and, if sufficient appropriations and resources are not provided, the legislature will consider adding a PRA. (Section 1, § 2415j; Section 2; Section 3.)

Updates to Data Broker Registry Headline Other Privacy Efforts

The VDPOSA may be the most notable privacy bill enacted in Vermont this year, but it is not the only one. Vermont also updated the state’s data broker registry (H.211), enacted a direct-to-consumer genetic testing law (H.639), and established a right to neural privacy (H.814).  

Data Brokers: Vermont is one of several states to create a data broker registry, alongside California, Connecticut (enacted this year), Oregon, and Texas. Effective January 1, 2027, H.211 significantly amends Vermont’s law. Key changes include—

• Re-defining “brokered personal information” (“BPI”) as “any information, including derived data and unique identifiers, that is linked or reasonably linkable, alone or in combination with other information, to an identified or identifiable individual or to a device that identifies, is linked to, or is reasonably linkable to one or more identified or identifiable individuals in a household.” The prior definition was narrower and limited to an enumerated list of identifiers, plus a catch-all. This definition, in contrast, is modeled on the broader definitions of “personal data” under many state comprehensive privacy laws. Additionally, the “publicly available information” exception to the definition of BPI has been broadened. The previous definition only excluded publicly available information “to the extent that it is related to a consumer’s business or profession.” That qualifier has been removed.
• Re-defining “data broker” by changing the “direct relationship” requirement. The core definition remains unchanged: a data broker is “a business . . . that knowingly collects and sells or licenses to third parties the [BPI] of a consumer with whom the business does not have a direct relationship.” Previously, the law provided examples of direct relationships, such as a “past or present . . . customer, client, subscriber, user, or registered user of the business’s goods or services.” The newly revised definition is instead modeled on California’s Delete Act regulations: a consumer must have “intentionally interacted with a business for the purpose of accessing, purchasing, using, requesting, or obtaining information about the business’s products or services,” and “[a] business is still a data broker and does not have a direct relationship with a consumer as to the [BPI] the business sells about the consumer that it collected outside of a first-party interaction with the consumer.” The revised definition also removed some of the law’s exceptions to data brokerage, such as “developing or maintaining third-party e-commerce or application platforms.”
• Adding a definition of “sale” that is consistent with the common definition under state comprehensive privacy laws: the exchange of BPI for monetary or other valuable consideration, subject to exceptions. 
• Imposing due diligence or ‘know your customers’–style requirements prior to disclosing BPI, such as: requiring “prospective users” of BPI to identify themselves, state their purposes for seeking the information, and certifying that the information will not be used for other purposes; and, prior to disclosing such information, verifying the prospective user’s identity, reviewing the stated purposes for using the information, and not disclosing the information where there are “reasonable grounds for believing that the information will be used to violate State or federal law or will not be used for the purposes stated by the user.” 
• Adding new security breach notice requirements specific to data brokers.
• Accelerating when a data broker must register, closing the timeframe to 30 days after a data broker first meets the relevant definition and then annually by July 1 of each year. 
• Increasing the registry fee and imposing a $20,000 bond. 
• Expanding the mandatory disclosures during registration. New information that a data broker must disclose includes whether the data broker: collects certain categories of personal information (e.g., precise geolocation data, immigration status); has sold or shared the consumers’ data, in the past year, with certain entities (e.g., foreign actors, state or federal government, “developer of a GenAI system or model”); “the URL of a page on the data broker’s website” that includes relevant information on consumer rights; and more. Notably, while this disclosure requirement references consumer rights to deletion, opting-out of the collection of BPI, and authorized agents’ ability to exercise an opt-out on the consumer’s behalf, the revised data broker law itself does not require that data brokers offer these rights. The newly enacted VDPOSA, however, will require data brokers to offer consumer rights if they are within the scope of that law. 
• Improving consumer access to the data broker registry in the form of a downloadable spreadsheet of registered data brokers.

Although earlier versions of H.211 would have added a California Delete Act–style accessible deletion mechanism, the final bill merely directs the Vermont Secretary of State to study the feasibility of establishing an accessible deletion mechanism. 

Genetic Testing: Vermont has become the third state this year—after South Dakota and Connecticut—to enact a law regulating direct-to-consumer genetic testing. The Vermont Genetic Information Privacy Act will go into effect on July 1, 2026. This law imposes notice and consent requirements for the collection and use of biological samples and genetic data, gives consumers rights of deletion and access, and prohibits certain disclosures or uses of genetic data. Violations of the law will constitute unfair and deceptive acts in commerce under 9 V.S.A. § 2453. This includes a private right of action, although consumers will have to provide written notice of an alleged violation to a direct-to-consumer genetic testing company or service provider prior to initiating a civil action and allow 30 days to cure the notice. The cure requirement will expire on June 30, 2028. The Attorney General has enforcement and rulemaking authority. 

Mental and Neural Privacy: Although the primary focus of H.814 is extending the duration and scope of the state’s Artificial Intelligence Advisory Council, this law also formally recognizes an individual right to “mental and neural data privacy.” This includes rights to “change an individual’s decision regarding neurotechnology,” to “be afforded protection from unauthorized neurotechnological alterations in mental functions critical to personality,” and to “be afforded protection from unauthorized neurotechnological alterations in mental functions critical to personality.” The law does not define key terms such as “neurotechnology,” nor are there specific mechanisms or business obligations attached to these new rights. The newly enacted VDPOSA includes neural data as a category of sensitive data, however, providing Vermont residents with actionable protections like opt-in consent requirements and mandatory data protection assessments.

* * *

Looking to get up to speed on the existing state comprehensive consumer privacy laws? Check out FPF’s 2025 report, Anatomy of a State Comprehensive Privacy Law: Charting the Legislative Landscape. 

Pictured: Vermont receiving its red star on the FPF “Privacy Patchwork” quilt.

Future of Privacy Forum Announces 2026 Career Achievement Award Recipients

WASHINGTON, D.C. — The Future of Privacy Forum, a global non-profit focused on data protection, AI, and emerging technologies, announced new recipients of its Career Achievement Award, recognizing exceptional leaders whose work has advanced privacy, responsible data governance, and AI leadership worldwide.

The 2026 recipients of the FPF Career Achievement Award are: 

• Anita L. Allen, one of the world’s foremost scholars of the law and ethics of data and technology, and Professor of Law at the University of Pennsylvania Penn Carey Law School; Henry R. Silverman Professor of Law and Professor of Philosophy, Emeritus   
• Nuala O’Connor, a leading voice in the responsible and ethical use of technology and data, a board member and senior advisor to companies and nonprofits, and the first Chief Privacy Officer of the Department of U.S. Homeland Security 
• Harriet Pearson, an esteemed technology governance and policy advisor and Managing Principal at Axia Advisory, former Cybersecurity Division Head, New York Department of Financial Services, former Partner, Hogan Lovells, first Chief Privacy Officer and Security Counsel, IBM Corporation.

Alan Raul, FPF’s Board Chair, and Jules Polonetsky, FPF CEO, were honored to introduce the awardees and presenters at the organization’s Advisory Board Annual Meeting in Washington, D.C.

“At a moment when conversations around privacy, AI, and digital governance are evolving faster than ever, these honorees represent the very best of principled leadership,” said Polonetsky. “Anita Allen, Nuala O’Connor, and Harriet Pearson have each made extraordinary contributions to advancing thoughtful, practical, and responsible approaches to data protection and emerging technologies. Their work has shaped the field in profound and lasting ways.”

“Each of this year’s honorees has helped define the modern privacy field through decades of leadership, scholarship, and public service,” added Raul. “Their work has influenced legal frameworks, corporate governance practices, academic thought, and public policy conversations around the world. FPF is proud to recognize their leadership.”

Anita L. Allen

Anita Allen is an internationally recognized scholar, philosopher, and legal theorist whose work has helped shape foundational conversations around privacy rights, ethics, civil liberties, and data governance. She is the Henry R. Silverman Professor of Law and Professor of Philosophy, Emeritus at the University of Pennsylvania Carey Law School and served as Vice Provost for Faculty. A prolific author, educator, and thought leader, Allen has examined issues of consent, accountability, equality, and ethical responsibility across both public and private life. Allen is also the author of more than 130 articles and chapters, as well as several books, including Unpopular Privacy: What Must We Hide.  She is an elected member of the National Academy of Medicine, the American Law Institute, the American Philosophical Society and a fellow of the American Academy of Arts and Sciences. Throughout her distinguished career, she has advised policymakers, data protection authorities, and institutions on questions at the intersection of privacy, technology, and democratic values, helping establish frameworks that continue to influence privacy law and ethics globally.

Nuala O’Connor

Nuala O’Connor is a globally recognized leader in the responsible use of data and technology in our daily lives. She currently serves as senior advisor on technology, privacy, and AI to multinational companies Maersk and Kekst CNC and is a board member at several nonprofit organizations including EqualAI, the Society for the Rule of Law, and the Future of Privacy Forum. Nuala has held technology and legal leadership roles at General Electric, Amazon, DoubleClick, and Walmart, and served as President and CEO of the Center for Democracy & Technology (CDT). She has held senior appointed positions in the U.S. Federal government, including serving as the country’s first statutory chief privacy officer, when she was appointed as CPO of the U.S. Department of Homeland Security. Throughout her career, O’Connor has been a leading voice advancing practical, human-centered approaches to emerging technology policy, responsible innovation, and digital governance.

Harriet Pearson

Harriet Pearson is one of the pioneers of modern corporate privacy leadership and cybersecurity governance. She is the founder of Axia Advisory LLC, where she consults on program design and governance, coaches senior leaders, and supports strategic communications and external engagement.  Pearson served for more than a decade as IBM’s first Chief Privacy Officer, helping establish one of the world’s most influential corporate privacy programs during a transformative period for global data governance and technology regulation. Until June 2024, Pearson was Executive Deputy Superintendent and head of the Cybersecurity Division at the New York Department of Financial Services where she led a comprehensive update of DFS’s cybersecurity regulation. Previously, Pearson was a Partner at Hogan Lovells where she founded and led the global cybersecurity practice for a decade starting in 2012. In addition to her corporate leadership, she has served in advisory and board roles across the technology, policy, and nonprofit sectors, helping organizations navigate complex questions around accountability, ethics, and responsible innovation. Pearson’s work has been at the leading edge of technology innovation and her work has led the implementation of privacy leadership as a core business and governance function at companies around the world.

FPF’s Annual Awards recognize individuals whose work has had a lasting impact on privacy protections, responsible innovation, and the advancement of ethical technology governance worldwide. 

Past FPF award winners are listed below.

Career Achievement Award

• Dale Skivington (2019) – former Chief Privacy Officer, Dell & Eastman Kodak
• Sandra Hughes – former privacy executive, Procter & Gamble
• Jane Horvath (2023) – Partner, Gibson Dunn, former Chief Privacy Officer, Apple
• Jim Halpert (2024) – General Counsel, White House Office of the National Cyber Director, Partner, DLA Piper
• Barbara Cosgrove (2025) – VP & Chief Privacy and Digital Trust Officer, Workday

Distinguished Public Service Award

• Jessica Rich (2017) – former Director, Bureau of Consumer Protection, Federal Trade Commission
• Helen Dixon (2019) – Data Protection Commissioner, Ireland
• Maneesha Mithal (2023) – Partner, Wilson Sonsini, former Associate Director, Division of Privacy & Identity Protection, Federal Trade Commission

Excellence in Career Award

• Patrice Ettinger (2024) – Chief Privacy Officer, Regeneron, former Chief Privacy Officer, Pfizer

Community Builder Award

• J. Trevor Hughes (2019) – President & CEO, IAPP

Outstanding Academic Scholarship Award

• Peter Swire (2019) – J.Z. Liang Chair, Georgia Tech School of Cybersecurity & Privacy Senior Counsel, Alston & Bird

Global Responsible AI Leadership Award

• Government of Singapore (2024)
• Brazil’s Autoridade Nacional de Proteção de Dados (ANPD) (2025)

Lifetime Achievement Award

• Julie Brill (2025) – Founder and Principal, Brill Strategies, former Commissioner, Federal Trade Commission, former Chief Privacy Officer, Corporate Vice President for Global Privacy, Safety and Regulatory Affairs, Microsoft

To learn more about the Future of Privacy Forum, visit fpf.org. 

##

Future of Privacy Forum Releases Comprehensive Report On Algorithmic Personalization in Youth Online Experiences

As policymakers continue to debate youth online safety regulations, a new FPF report assesses the role of data-driven personalization and its implications for emerging policy and product design

WASHINGTON, D.C. — (June 10, 2026) — The Future of Privacy Forum — a global non-profit focused on data protection, AI, and emerging technologies —today released Personalization and Youth Online: Assessing Benefits, Risks, and Safeguards, a comprehensive report examining the role of algorithmic personalization in young people’s digital lives and its implications for emerging policy and product design.

Personalization—the use of personal data to tailor content and services to individual users—sits at the center of today’s most contested debates about youth and digital technology.  It appears in nearly every digital service young people use—influencing the content they see, how information is ranked, which recommendations are surfaced, and what prompts or ads are presented. While it offers a host of functional and protective benefits, the practice has also become a recurring regulatory focus of efforts to protect minors from online harms, including by targeting the specific product features and design elements through which personalization operates.

“Personalization is an incredibly nuanced topic; the same mechanisms that may be associated with risks may also enable benefits such as adaptive learning, community discovery, and age-appropriate protections that make the online experience safer for young people,” said Daniel Hales, Policy Counsel for the Future of Privacy Forum and co-author of the report. “Understanding the tradeoffs that come with any risk mitigation strategy is critical, so you don’t inadvertently limit the benefits of personalization along with the harms.”

In an effort to help policymakers and companies effectively evaluate these tradeoffs, the report examines the intersection of personalization and youth online experiences in five parts: key definitions and common use cases; key benefits of personalization in youth online experiences; common risks of harm; an assessment of more than a dozen emerging mitigation proposals and the associated tradeoffs; and an analysis of the emerging policy and regulatory landscape.

Examples of the mitigation strategies examined in the report include:

• Alternative approaches to curating content, such as feeds that display content chronologically, by popularity, or by categories and preferences a user has explicitly selected. While these approaches may reduce data-intensive curation, the report notes that they may have unintended effects, including the potential to inadvertently amplify spam and misinformation, and limit platforms’ ability to suppress age-inappropriate content.
• Data minimization rules applied without reference to the purpose of how the data is being used may curtail beneficial personalization alongside higher-risk applications. Purpose-based limitations, which restrict data use according to processing context rather than data category alone, may better preserve the protective uses of personalization while constraining its riskier applications.
• Risk mitigations related to conversational AI chatbots, including those designed to facilitate transparency, consent, responsible data practices, and safety-by-design. Each of these comes with its own tradeoffs; for example, on-device processing of sensitive data protects privacy, but can reduce model quality, which may ultimately push users toward less protective alternatives. Effective safeguards for AI personalization will require careful attention to which constraints reduce overall risk and which simply shift it. 
• Safety and well-being measures, including parental control and oversight, increasing digital literacy, measures to improve digital well-being, and control over personal data use. Many of these tactics will be most effective when paired with other measures; for example, increased digital literacy and making default wellbeing features like time-use limits and sleep and nighttime protections do not address the underlying design features that generate risk.

“Effective regulation must account for some precise questions: how personalization is implemented, what data it uses, and what purpose it serves,” said Holly Hawkins, Director for Youth Policy for the Future of Privacy Forum and the other co-author of the report. “While no single mitigation strategy is going to effectively address the full range of risks, we know that the most effective approaches share a common characteristic: they take a risk-proportionate approach to limiting higher-impact potential harms while preserving the functional and protective benefits of personalization for young people. We hope that this report can serve as a valuable resource to policymakers and companies who are attempting to find this delicate balance.”

The report’s release follows a growing interest from policymakers at both the federal and state levels in regulating personalization practices as part of efforts to strengthen protections for youth online. Both New York and California have passed laws prohibiting online services from providing algorithmically curated feeds to minors without parental consent; South Carolina and Nebraska’s age-appropriate design code laws require services to offer an opt-out of personalized recommender systems.

The full report, including an appendix tracking a sample of enacted and proposed youth online safety laws worldwide that address personalization, is available here. 

To learn more about the Future of Privacy Forum, visit fpf.org. 

###

Frontier AI Goes Federal: How the Great American AI Act Compares to State Laws

Introduction

It has been an unusually active few weeks for AI safety policy. Following a new frontier model safety bill passed in Illinois, and a White House executive order on AI security, Rep. Jay Obernolte (R-CA) and Rep. Lori Trahan (D-MA) released a bipartisan discussion draft for the Great American AI Act of 2026, adding another major federal proposal to the rapidly developing frontier AI landscape.

The draft is broad, covering issues ranging from workforce development and AI literacy to cybersecurity and international standards. But for many AI developers and deployers, the most important provisions are those focused on frontier model regulation. The draft would create requirements related to frontier AI transparency, critical safety incident reporting, employee whistleblower protections, and independent verification organizations. It would also include a three-year preemption clause restricting state laws that specifically regulate AI model development.

This blog highlights four key takeaways of the discussion draft: 

1. The draft is one of the first bipartisan attempts in Congress to address both frontier model safety and preemption of AI. These aspects make it a notable legislative effort, even if its prospects are uncertain. 
2. The draft incorporates many of the frontier model safety provisions in existing state laws but also has key distinctions. Compared to recent state frontier AI laws in California, New York, and Illinois (pending signature), the bill makes some important adjustments, like adding a revenue threshold for “frontier developers,” modifying the definition of “critical safety incident,” and utilizing a different penalty structure. 
3. The draft brings the preemption debate back into the federal AI policy conversation. It includes a three-year preemption clause focused on state laws that specifically regulate AI model development. 
4. The draft also reaches beyond frontier model safety. Other notable provisions include a study content moderation, a federal voluntary model testing program, and disclosure requirements for AI-related mass layoffs.

The Act Enters the AI Safety and Preemption Debate 

Amidst a crowded but unsettled federal AI policy landscape, the Great American AI Act is notable for its regulatory focus, bipartisan backing, and what it may signal for federal AI governance ahead. Though Congress has introduced no shortage of AI bills, there has been limited movement toward enacted legislation. Congress has considered sector-specific bills on chatbots¹, regulatory sandboxes, defense, elections, and financial scams, as well as broader proposals aimed at establishing a national AI framework like Sen. Blackburn’s (R-TN) TRUMP AMERICA AI Act.

Amid this activity, the Great American AI Act steps into two of the most active and contested AI policy debates: frontier model safety and federal preemption. On AI safety, the draft follows a period of growing attention to frontier model oversight at both the state and federal levels, where policymakers are trying to balance concerns about catastrophic risks and national security with concerns that overly burdensome requirements could slow AI innovation or weaken U.S. competitiveness.

On preemption, the draft arrives less than a year after Congress rejected a much broader effort to pause state AI regulation. In July 2025, the Senate voted 99-1 to remove a proposed 10-year moratorium on state AI laws from the budget reconciliation package. By contrast, the Great American AI Act includes a narrower three-year preemption provision focused on state laws that specifically regulate AI model development.

The draft is also significant because of both who introduced it and how they introduced it. Other federal AI bills have addressed AI safety or included preemption language, but this proposal comes from bipartisan sponsors who have been closely involved in federal AI policy². In introducing the draft, the sponsors emphasized that “policy for a technology this transformative can only be built to last if it’s written by both parties.” Just as importantly, they have framed the draft as a starting point rather than a final product, describing it as “the start of a serious national conversation.” Whether the bill advances in its current form remains uncertain, especially given the sensitivity of federal preemption and the range of issues addressed. Even if the draft changes, it may help shape the terms of future federal debates over AI safety. At a minimum, it is another sign that Congress is increasingly focused on how to govern AI systems and how federal rules should interact with the fast-moving state AI landscape.

Frontier AI Requirements: Where the Draft Aligns with State Laws

The draft’s frontier model provisions are not starting from scratch. They closely track the framework emerging from California’s SB 53, New York’s RAISE Act, and Illinois’ SB 315 (awaiting signature), including requirements for frontier AI frameworks, transparency reports, safety incident reporting, enforcement, and whistleblower protections. But the federal draft makes some important adjustments, including a revenue threshold for “frontier developers,” a definition of “critical safety incident” that is broader in some respects and narrower in others, and a federal enforcement structure with penalties of up to $1 million per day.

Key Definitions

The Draft largely aligns with recent state frontier AI laws’ definitions of “large frontier developer,” “frontier model” (encompassing foundation models trained using more than 10^26 computational operations), and “catastrophic risk” (covering certain risks of death, serious injury, or major property damage arising from frontier models). For more background on SB 53 and the RAISE Act, see FPF’s prior analysis. There are, however, a few notable differences. 

First, unlike the state laws, the Draft requires “frontier developers” to have a gross revenue exceeding $50 million. The state laws only include a gross revenue qualifier in the definition of “large frontier developer.”

Second, the draft’s definition of “catastrophic risk” does not  specify that the death, serious injury, or property damage must arise from a “single incident,” as in the state laws.

Finally, the draft’s definition of “critical safety incident” differs from the state laws by not mandating actual harm occur (such as death, bodily injury, or property loss), nor does it include scenarios where the model uses deceptive techniques to evade developer controls or monitoring.

Core Frontier Model Obligations

The draft would impose several obligations on frontier developers and large frontier developers that also largely mirror the recent state frontier AI laws. The main requirements include:

• Frontier AI Framework: Large frontier developers would need to write and publicly post a frontier AI framework addressing catastrophic-risk thresholds and mitigations, model weight cybersecurity, internal governance, and decisions about internal use and deployment. Developers would need to review the framework at least annually and publish any material modification within 30 days. These requirements closely track the state laws.
• Transparency Reports: Before, or concurrently with, deploying a new frontier model or a model with a substantial modification, frontier developers would need to publish a transparency report covering information such as release date, supported languages, output modalities, intended uses, restrictions, risk assessments, third-party involvement, and mitigation steps. These requirements largely align with recent state frontier AI laws.
• Reporting Mechanisms: The draft would require confidential reporting mechanisms for critical safety incidents and for catastrophic risks arising from a large frontier developer’s own use of its frontier models, including internal use and internally deployed models. This is similar to the state laws in allowing both frontier developers and members of the public to report critical safety incidents, but the draft adds a separate mechanism for reporting catastrophic risks from developers’ own model use.
• Critical Safety Incident Disclosures: Frontier developers would need to report critical safety incidents to the Center for AI Standards and Innovation (CASI) within 15 days of discovery, or within 24 hours if the incident poses an imminent risk of death or serious injury. This mirrors California’s SB 53, while New York and Illinois require a shorter 72-hour reporting window.
• Enforcement: Violations can result in fines of up to $1 million per violation, with each day treated as a separate violation. Federal and state attorneys general may also seek injunctions. This penalty structure may be less stringent than the state laws: SB 53 allows penalties of up to $1 million per violation, while the RAISE Act and SB 315 allow penalties of up to $3 million for subsequent violations. However, unlike the draft, those laws do not clearly cap penalties for a continuing violation at $1 million per day, meaning multiple violations could potentially be enforced in a single day.
• Definitional Updates and Rulemaking: Beginning by January 1, 2028, the CAISI Director would need to issue annual recommendations on whether key definitions should be updated. The Secretary of Commerce would also have authority to adopt rules implementing the provisions, including criteria for determining when model modifications are “substantial” or “material.” California’s SB 53 includes a similar definitional update requirement, while New York’s RAISE Act similarly provides rulemaking authority.
• Employee Whistleblower Protections: In a separate section, the draft would prohibit AI companies from retaliating against employees for lawfully reporting violations of federal AI laws and would provide remedies for employees who experience retaliation. This provision is similar to whistleblower protections in California’s SB 53 and Illinois’ SB 315, though those laws go further by requiring frontier developers to provide covered employees with clear notice of their rights and access to a reasonable internal process for anonymously disclosing information.

Taken together, the draft would bring much of the emerging state frontier AI framework into federal law. The broad architecture is familiar, but the federal draft contains key distinctions. It also does not carry over every state-law mechanism, most notably the frontier developer disclosure programs in the RAISE Act and SB 315, which require large frontier developers to maintain current filings with state agencies on ownership and business information. 

Independent Verification Organizations and Audits

The draft would also put independent auditors at the center of its frontier AI framework. Within one year of enactment, and every six months thereafter, large frontier developers would need to retain a licensed independent verification organization (IVO) to verify compliance with the draft and assess whether the developer’s risk mitigation efforts are adequate to address catastrophic risks. 

The provision reflects a broader trend toward independent evaluation of frontier models. Illinois SB 315 would also require large frontier developers to undergo independent third-party audits, but it does not create a licensing and oversight system for IVOs like the one proposed in the draft. However, other states, like Virginia, have begun exploring this model. Virginia enacted SB 384 this year, which directs the Joint Commission on Technology and Science to study the future development of an IVO framework (after earlier versions of the bill would have created an IVO licensing structure akin to this draft).

These provisions would require:

• Access: Large frontier developers would have to provide IVOs with timely access to unredacted materials and other information reasonably necessary to conduct audits and assessments. Developers could impose reasonable security protocols and access limitations to protect trade secrets and confidential business information.
• Audit and Assessment Content: IVO reports would have to describe the scope, time period covered, materials reviewed, methods used, and any limits on the assessment. They would also assess the developer’s compliance efforts; the adequacy of its frontier AI framework, and risk-monitoring; any “failure, deficiency, or material weakness;” and internal controls.
• Ad Hoc Audits and Assessments: The Director of CASI could require additional audits when necessary to verify compliance, validate prior findings, or monitor significant changes in risk, including after a critical safety incident or substantial modification.
• Supplemental Reports: IVOs could also need to submit supplemental reports when new information, unexpected model capabilities, or major model changes call into question an earlier finding that the developer was adequately managing catastrophic risks.

Additionally, the Bill would create a federal licensing and oversight system for the auditors themselves, making IVOs a potential new layer of AI governance.

Together, these proposals suggest that policymakers are increasingly looking to independent verification as a middle ground between company self-assessment and direct government review. The Great American AI Act would take that idea further by specifying who can serve as an independent verifier, what they must review, and how their findings may be shared with enforcement authorities.

Federal Preemption and the State AI Landscape

One of the draft’s most consequential provisions is its three-year preemption clause. The draft would prohibit states from establishing, continuing in effect, or enforcing any law or regulation that specifically regulates the development of an AI model (emphasis added). It would not preempt laws of general applicability, state authority granted under the draft, or laws governing post-deployment activities, including the implementation, distribution, offering, or use of AI systems, products, or services.

This approach differs significantly from the AI moratorium Congress considered last year as part of the One Big Beautiful Bill Act. That proposal would have broadly restricted states from enforcing laws or regulations that “limit,” “restrict,” or “regulate” AI models, AI systems, or automated decision systems, while allowing laws viewed as “facilitating” AI. It also evolved from a proposed ten-year pause to a five-year compromise before ultimately being removed from the package. By contrast, the Great American AI Act uses narrower language: it applies only to laws that specifically regulate AI model development and sunsets after three years.

However, as always, preemption would be complicated. Many state AI laws do not neatly separate development from deployment. They may impose obligations on developers and deployers, or require pre-deployment documentation or risk assessments for systems that are ultimately used in employment or other high-impact contexts.

As drafted, the draft’s preemptive effect would likely be greatest for state laws focused on frontier model development, such as California’s SB 53, New York’s RAISE Act, and Illinois SB 315. But the clause could also create uncertainty for other AI laws. For example, a chatbot law focused on how companies offer chatbots to the public would likely be easier to preserve under the post-deployment carveout. But if a law requires changes to how a model is trained, fine-tuned, tested, or designed before release, the line between deployment and development may be harder to draw.

The result is a preemption provision that is far narrower than last year’s proposed 10-year moratorium, but still raises significant interpretive questions. The answer to those questions is likely to have an impact on both developers and deployers, and would likely depend on how regulators and courts characterize specific laws and their requirements.

Other Key Provisions

Although the frontier model provisions are the most significant for most AI developers, the draft reaches well beyond AI safety: It also includes several provisions focused on the broader AI ecosystem, including workforce development and displacement. This section highlights three additional provisions worth watching: a study on government engagement with AI platforms, new federal AI testbeds and voluntary model testing, and disclosure requirements for AI-related mass layoffs.

• Study on Protecting Free Speech: The draft would direct the Secretary of Commerce to study how the government encourages or pressures AI companies on content moderation, information prioritization, and output generation. The report must examine legal frameworks governing federal agency interactions with AI platforms and recommend legislation providing individuals with redress against unlawful government censorship, including transparency and oversight mechanisms. While this section is only a study requirement, it signals congressional interest in how the government communicates with AI companies about content moderation and model outputs. For AI companies, this study may lead to future regulation on transparency and government oversight.
• AI Testbeds: The draft would direct the establishment of an AI testbed program for testing, evaluating, and assessing AI systems, including automated evaluations, security vulnerability assessments, and computational resource assessments. It would also establish a voluntary foundation model testing program for vendors of foundation models, AI virtual agents, and robots that incorporate foundation models. This provision parallels recent federal activity around voluntary AI testing and evaluation, including the White House’s June 2026 executive order establishing a voluntary framework for certain frontier AI developers to share models with the federal government before public release for national security and cybersecurity assessment.
• Disclosures for AI Layoffs: The draft would require employers to provide 60 days’ advance notice when AI is a “substantial factor” in a mass layoff, including information on the AI involved, the estimated share of job losses attributable to AI, and pre-layoff upskilling or retraining efforts. A similar requirement was recently enacted in Connecticut’s SB 5, which requires employers issuing mass layoff notices to tell the state’s Labor Department whether the layoffs are related to the employer’s use of AI. This provision reflects growing interest in how AI may affect workforce displacement and whether existing worker-notice frameworks should account for AI-related job losses.

Conclusion

The Great American AI Act is still a discussion draft; whether it will advance in its current form remains uncertain. Federal AI legislation has faced a difficult path in Congress and the draft touches on issues, especially frontier model safety and preemption, that are likely to generate significant debate.

But the draft is still an important marker in the federal AI conversation. It shows that Congress is continuing to consider how to regulate frontier model development, how to structure independent oversight, and how federal rules should interact with the growing number of state AI laws. Even if this bill changes substantially, these questions are unlikely to go away.

In the meantime, states are likely to continue to test different approaches to AI regulation, including laws that regulate frontier model development, deployment, or both. That could make the draft’s preemption language especially important to watch. The next phase of AI policymaking could be defined not only by the rules new proposals would set, but also by the coalitions advancing them and the venues where they move forward: Congress, the Executive Branch, the states, or all three.

1. Five major federal chatbot proposals have been introduced: the CHAT Act (S.2714/H.R. 7218), GUARD Act (S.3062/HB 8623), SAFE BOTs Act (H.R. 6489), Youth AI Privacy Act (S. 4199), and the CHATBOT Act (S 4407). ↩︎
2. Rep. Obernolte (R-CA) previously co-chaired the bipartisan House Task Force on Artificial Intelligence, which released its final report in December 2024. Rep. Trahan (D-MA), a member of the House Energy and Commerce Committee’s Innovation, Data, and Commerce Subcommittee, has also been active on technology accountability issues. ↩︎

Privacy Becomes You, Bayou State: A Look at the Louisiana Data Privacy Act

Louisiana has become the 22nd U.S. state to enact a comprehensive consumer privacy law—and the third this year following Oklahoma and Alabama—after Governor Landry signed the Louisiana Data Privacy Act (LDPA) (SB 386) on May 29. Overall, this is a fairly standard state privacy law that follows the Washington Privacy Act framework apart from the law’s CCPA-style applicability thresholds. The law will go into effect on January 1, 2027. This blog post covers the LDPA’s scope, consumer rights, business obligations, and enforcement. 

Scope

Applicability: Like other comprehensive privacy laws based on the Washington Privacy Act (WPA) framework, this law regulates controllers’ and processors’ collection and use of personal data.

Departing from the common WPA framework, this law’s applicability thresholds are modeled on those under the California Consumer Privacy Act (CCPA). The LDPA only applies to a person or entity doing business in Louisiana that either—

1. Has annual gross revenues exceeding $25 million; 
2. Annually buys, “receives for the business’s commercial purposes,” sells, or shares for commercial purposes the “personal information” of at least 75,000 consumers, households, or devices; or 
3. Derives 50% or more of its annual revenues from selling consumers’ “personal information.”

Not only do these thresholds reflect those under the CCPA, they also use the undefined term “personal information” (which is used in the CCPA) rather than the defined term “personal data” used throughout the LDPA. One unique aspect of these applicability thresholds is that prong (2) adds the criteria “receives for the business’s commercial purposes,” which is not present in the CCPA’s text although that law defines “commercial purpose” and uses the term in other contexts. (§ 1780.2(A).)

The LDPA includes broad entity- and data-level exemptions, including for—

• State agencies or political subdivisions;
• Financial institutions, affiliates, or data subject to the GLBA; 
• Covered entities, business associates, and protected health information governed by HIPAA; 
• Health records;
• Information included in a limited data set maintained as required under 45 C.F.R. 164.514(e);  
• Nonprofits, including political organizations; 
• Institutions of higher education; 
• Information collected, used, or disclosed by a consumer reporting agency or furnisher to the extent regulated by and authorized under FCRA; 
• Personal data collected, processed, sold, or disclosed in compliance with the DPPA; 
• Personal data regulated under FERPA; and more. (§ 1780.2(B)-(C).)

Key Definitions: Personal data is defined consistently with other state laws as information that is linked or reasonably linkable to an identified or identifiable individual, and it does not include deidentified data or publicly available information. Sensitive data includes: personal data revealing racial or ethnic origin, religious beliefs, mental or physical health diagnosis, sexuality, or citizenship or immigration status; genetic or biometric data processed for uniquely identifying an individual; personal data collected from a known child (under 13); and precise geolocation data (within a radius of 1,750’). This definition is narrower than in many of the newer state laws, which often include other categories such as consumer health data, neural data, or status as a victim of a crime. (§ 1780.1.)

This law includes many of the key definitions associated with the Connecticut-model of state laws. For example: the definition of “biometric data” includes data generated from a photograph or video or audio recording if generated to identify a specific individual; “dark patterns” are defined and prohibited for obtaining consent; and “sale” is defined broadly to include exchanges of personal data for “other valuable consideration” apart from monetary consideration. (§ 1780.1.)

Consumer Rights

Consumers will have the standard rights to: confirm whether their personal data is being processed; access their personal data; correct inaccuracies in their personal data; have their personal data deleted; obtain a copy of their personal data in a portable format (if available in a digital format); and opt-out of the processing of their personal data for the purposes of targeted advertising, the sale of personal data, and profiling in furtherance of a decision that produces a legal or similarly significant effect concerning the consumer. Like the laws in Tennessee and Alabama, these consumer rights (including the opt-out right) do not apply to pseudonymous data if the controller is able to demonstrate that information necessary to identify the consumer is kept separately and subject to effective technical and organizational controls that prevent the controller from accessing the information.  (§ 1780.3(A), 1780.4(O),)

Controllers must respond to consumer rights requests within 45 days, which can be extended an additional 45 days if necessary so long as the consumer is informed of the extension and the reason. If a controller declines to act on a consumer request, then it must inform the consumer of the decision, the justification, and how to appeal the decision. A controller is not required to comply with a rights request that it cannot authenticate, and that the authentication requirement extends to the consumer opt-outs as well. Some states, like Connecticut, provide that a controller does not need to authenticate an opt-out request but may deny an opt-out request if it has a good faith, reasonable and documented belief that the request is fraudulent. (§ 1780.3(B).)

In another departure from the Connecticut-style suite of state laws, the LDPA does not appear to require a controller to provide consumers with a mechanism to revoke previously given consent. 

What about Agents and OOPS? A consumer will be able to designate another person to serve as the consumer’s authorized agent to opt-out of the processing of consumer’s personal data for targeted advertising or the sale of personal data. Although the law does not reference opt-out preference signals (OOPS) or universal opt-out mechanisms (UOOM), it does provide that a consumer can “designate an authorized agent using a technology, including . . . a global setting on an electronic device,” that allows the consumer to indicate the consumer’s intent to opt out of the processing for targeted advertising, for sale of personal data, or both.” Additionally, a “technology” described in the subsection may not “unfairly disadvantage another controller,” make use of a default setting (instead requiring “an affirmative, freely given, and unambiguous choice” by the consumer), and be consumer-friendly and easy to use. These are the common requirements for an OOPS under the state comprehensive privacy laws. 

The use of a technologically-designated authorized agent by a consumer could be limited due to several exceptions under the law. A controller is not required to comply with an opt-out request from an authorized agent if: the authorized agent does not communicate the request in a clear and unambiguous manner; the controller cannot verify (with reasonable effort) that the consumer is a resident of Louisiana; the controller does not possess the ability to process the request; or the controller “does not process similar or identical requests the controller receives from consumers for the purpose of complying with similar or identical laws or regulations of another state.”(§ 1780.3(E)(5)-(6).) 

Business Obligations

Consistent with most of the state privacy laws, controllers and processors are subject to an enumerated list of duties under the law—including transparency, data minimization, data security, non-retaliation, oversight of processors, data protection assessments, and children-specific protections—as well as a list of broad exceptions. 

Transparency: Controllers must provide consumers with a “reasonably accessible and clear privacy notice” including information such as categories of personal data processed, processing purposes, how consumers can exercise their data rights, categories of personal data sold to third parties, and categories of third parties to whom data is sold. If the controller sells personal data, processes personal data for targeted advertising, sells sensitive data, or sells biometric data, there are additional notices that must be provided in the privacy notice (e.g., “NOTICE: We may sell your sensitive data”). (§ 1780.4(B).)

Data Minimization: The LDPA includes common procedural data minimization and purpose limitation restrictions. A controller must— 

• “[L]imit the collection of personal data to what is adequate, relevant, and reasonably necessary in relation to the purposes for which that personal data is processed, as disclosed to the consumer” 
• Obtain the consumer’s consent to “process personal data for a purpose that is neither reasonably necessary to nor compatible with the disclosed purpose for which the personal data is processed, as disclosed to the consumer”; and 
• Obtain the consumer’s consent to process the consumer’s sensitive data. (§ 1780.4(A).)

Data Security: A controller must establish, implement, and maintain reasonable administrative, technical, and physical data security practices that are appropriate to the volume and nature of the data. (§ 1780.4(A)(1)(b).)

Anti-discrimination and Non-retaliation: Controllers cannot process personal data in violation of state and federal laws that prohibit unlawful discrimination against consumers. Controllers also may not deny goods or services, charge different prices or rates for goods or service, or provide a different level of quality of goods or services to the consumer as retaliation for a consumer exercising any of their rights under the LDPA, subject to exceptions (e.g., if the data is necessary to provide a service or processed in connection with a bona fide loyalty program). (§ 1780.4(A).)

Processors: Processors must adhere to the instructions of a controller and assist the controller in complying with the controller’s duties or requirements under the law. Whether a person is acting as a controller or processor is a fact-based determination depending on context, but a processor remains a processor if they are adhering to a controller’s instructions with respect to a specific processing activity. There must be a valid contract in place between the controller and processor that meets statutory criteria (e.g., clear instructions for processing, deleting or returning personal data after the service is concluded). (§ 1780.4(D).)

Children’s Privacy: Consistent with most other state comprehensive privacy laws, the LDPA includes protections for children’s personal data and provisions that address COPPA compliance. “Sensitive data” includes personal data collected from a known child, and a controller must process the sensitive data of a known child in accordance with COPPA. Parents are able to exercise consumer rights on behalf of a child whose personal data is processed. Controllers and processors that comply with the verifiable parental consent requirements of COPPA are deemed to be in compliance with any requirement to obtain parental consent under the LDPA. In contrast to many of the newer state comprehensive privacy laws, the LDPA does not include opt-in rights for teenagers with respect to targeted advertising or the sale of personal data. (§§ 1780.1, 1780.2(E), 1780.3(A) & 1780.4(A).)

Data Protection Assessments: Controllers must conduct and document a data protection assessment for processing activities that present a heightened risk of harm to consumers, including processing personal data for targeted advertising, selling personal data, processing personal data for profiling that presents a reasonably foreseeable risk of substantial injury to consumers, and processing sensitive data. A controller must make a data protection assessment available to the Louisiana attorney general if requested in a civil investigative demand (although that requirement includes a cross-reference to a non-existent subsection of the law). (§ 1780.4(E).)

Exceptions: This bill includes a number of common exceptions, providing that nothing in the law shall be construed to limit a controller’s or processor’s ability to: comply with state, federal, or local laws or regulations; comply with regulatory inquiries or investigations; provide a specifically requested product or service; engage in public or peer-reviewed research in the public interest adhering to relevant safeguards; cooperating with law enforcement agencies; internal use of data for conducting research, effectuating a product recall, identifying and repairing technical errors, performing internal operations reasonably aligned with consumers’ expectations; and more. (§ 1780.4(G)-(I).) 

Miscellaneous: This law includes one unique provision related to the sale of sensitive data. Section 1780.4(P) provides that “[a] person or entity described by R.S. 51:1780.2(A)(3) may not engage in the sale of personal data that is sensitive without receiving prior consent from the consumer,” and violation of that requirement subjects a person to a penalty under the law. The cross-reference is to the applicability threshold for a person or entity that does business in the state and that derives fifty percent or more of its annual revenues from selling consumers’ “personal information.” This is an ambiguous requirement. An entity meeting that threshold would already be under the requirement to obtain consent prior to processing sensitive data, which includes selling data, so it is not clear that this is an added responsibility, unless it is meant to apply more broadly. But there is no other language in the requirement suggesting that it would apply notwithstanding the law’s broad entity-level exemptions.

Enforcement

The Louisiana Attorney General will enforce the LDPA and violations will constitute unfair and deceptive trade practices under Louisiana’s Unfair Trade Practices and Consumer Protection Law. Notably, the private rights of action under the unfair trade practices law do not extend to violations of the LDPA. For the first six months of enforcement (January 1, 2027 to July 31, 2027), the attorney general must give persons notice of alleged violations and at least 30 days to cure those violations prior to initiating an investigation. 

The attorney general is required to post online information regarding controllers’ and processors’ responsibilities and consumer rights under the law, and money received from enforcement actions will go towards funding the attorney general’s consumer protection efforts or promoting consumer protection and education. 

* * *

Looking to get up to speed on the existing state comprehensive consumer privacy laws? Check out FPF’s 2025 report, Anatomy of a State Comprehensive Privacy Law: Charting the Legislative Landscape. 

Pictured: Louisiana receiving its star on the FPF “Privacy Patchwork” quilt.

Comparing Enacted App Store Accountability Acts

On May 28, 2026, the 5th Circuit granted a stay on the preliminary injunction blocking enforcement of Texas’s App Store Accountability Act (ASAA)—meaning the law is now in effect while litigation on the merits continues. In 2025, Utah, Texas, and Louisiana enacted App Store Accountability Acts (ASAAs) which impose novel and significant age assurance obligations on app store providers and app developers. These laws require account holder age verification and parental consent for minors at the app store level, with age band and consent data transmitted between the two parties via bespoke “age signals.” Texas’s law was the first scheduled to go into effect on January 1, 2026. Before the law took effect, however, two groups filed suit challenging its constitutionality on First Amendment grounds. In December 2025, a federal judge issued a preliminary injunction blocking the law, finding it more likely than not an unconstitutional content-based regulation. Attorney General Paxton quickly appealed this decision to the 5th Circuit seeking a stay on the injunction, which was subsequently granted last week. 

The uncertain constitutional outlook of ASAAs puts compliance teams in a difficult position. Moving slowly risks liability if the laws survive legal challenge; moving quickly risks sunken compliance costs if ASAAs are ultimately struck down—costs that could have otherwise been diverted to other important trust and safety priorities. While Utah’s, Texas’s, and Louisiana’s laws impose broadly similar obligations on app store providers and developers, important distinctions will shape how companies ultimately achieve compliance. Amendments to Utah’s and Louisiana’s laws in 2026 further shape the developing obligations in each state. FPF created a comparison chart detailing the key terms, scope, and core obligations of these laws, including changes incorporated by 2026 amendments. The chart also includes an Appendix detailing current information about Developer APIs released publicly by app stores to aid developers as these requirements go into effect. 

Key takeaways from this resource include:

• Scope: These laws apply broadly to apps publicly available to consumers for download in an app store with very narrow exceptions, such as for settings apps or essential device drivers, and no wholesale exemptions from application. Uniquely, Louisiana also applies to apps that are available for download from an app store onto connected devices.
• Obligations: These laws require app store providers to implement age verification and obtain parental consent for minor account holders making purchases, downloads, in-app transactions, or significant account changes. Utah and Louisiana extend this requirement to a minor’s first-time access to many pre-loaded apps. Developers must request and receive age signals from app stores, use that data to manage minor access, and implement appropriate safety features. Utah and Louisiana add notable nuances for developers, such as allowing them to request that app stores block minor access to their apps entirely, or permitting reliance on a primary account holder’s age data for “family account apps.”
• Enforcement: Private rights of action (PRAs) are a trending enforcement mechanism in ASAA models. While Utah relies solely on a PRA for enforcement, Texas’s law only allows a PRA through incorporated references to the Deceptive Trade Practices Act (DTPA). Louisiana is the only law that  does not include a PRA.

The ASAA trend has continued into 2026 with Alabama enacting a new ASAA law back in February. Alabama’s law largely tracks with the ASAA models in Utah and Louisiana following their 2026 amendments. As ASAA legislation and litigation continues to develop, the 5th Circuit’s stay on injunction means that compliance teams must figure out how to navigate this growing thicket of app store age signals and online safety requirements in the meantime.