FPF Files Comments for the FTC Health Breach Notification Rule Addressing Specific Definitions and Clarity of Scope
On August 8th, the Future of Privacy Forum (FPF) filed comments with the U.S. Federal Trade Commission (the Commission) regarding the Notice of Proposed Rulemaking (NPRM) to clarify the scope and application of the Health Breach Notification Rule (HBNR).
The HBNR was promulgated in 2009 as part of the American Recovery and Reinvestment Act as a breach of security rule. Recent complaints brought by the Commission, GoodRx and Easy Healthcare, were the inaugural and second application of the HBNR and indicated a novel range of alleged privacy breaches rather than traditional security breaches. The cases indicated a shift in the interpretation of “breach of security” by the Commission that drew many proto-typical practices into scope. The NPRM seeks to clarify this broadened scope which has amalgamated traditional breaches of security with nascent breaches of privacy. To draw out and address key issues in the NPRM and the Commission’s considerations, we recommended that the Commission consider the nuance of definitions and address the complexities of breach by specifically:
Define a Standard for Identifiability for “PHR identifiable health data” to Clearly Expand Protections for a Broad Spectrum of Personal Information
Define “Relates to” to Include the Creation of Health-Related Inferences from a Wide Range of Routine Commercial Datasets, While Establishing Clear Obligations for Businesses
Establish Clear Guidelines for Intentional Data Sharing that Does Not Require Affirmative Consent
Ensure that the Rule Contains “Good Faith” Exceptions for Merely Technical Violations
Further Define “Breach of Security” to Clarify Where the Commission May Take Enforcement Action
FPF’s full comments to the Commission are available here.
FPF Releases Generative AI Internal Policy Checklist To Guide Development of Policies to Promote Responsible Employee Use of Generative AI Tools
Today, the Future of Privacy Forum (FPF) releases the Generative AI for Organizational Use: Internal Policy Checklist. With the proliferation of employee use of generative AI tools, this checklist provides organizations with a powerful tool to help revise their internal policies and procedures to ensure that employees are using generative AI in a way that mitigates data, security, and privacy risks, respects intellectual property rights, and preserves consumer trust.
The Checklist draws from a series of consultations with practitioners and experts from over 30 cross-sector companies and organizations to understand current and anticipatory employee use of generative AI tools, benefits and harms, AI governance, and measures taken to protect company data and infrastructure. The conversations focused on any generative AI guidelines, policies, and procedures that had been implemented to govern employees’ use of generative AI tools.
From those discussions, we learned that organizations have broadly varied use cases for generative AI and, therefore, significant variation in generative AI policies. Some organizations have enacted outright bans for generative AI tools without prior approval, while others have created restrictions for the use of generative AI, and still, others have yet to develop express policies and procedures on employee use of generative AI. The Internal Policy Checklist for Generative AI is intended to serve as a guidance document no matter what stage of the process an organization is in. It may be used as a starting point to help kick off the development of internal generative AI policies or as a final check to ensure an organization has provided comprehensive and robust guidelines for their teams.
“It is imperative that both organizations and their employees understand the benefits and risks of generative AI tools, and that organizations have appropriate safeguards in place to support responsive and ethical use,” said Amber Ezzell, AI policy counsel at FPF and author of the checklist. “Employee use of generative AI tools is inevitable and may bring new and unexpected benefits to employers as employees find ways to be more productive and creative in even the most mundane tasks. Developing thoughtful generative AI policies is essential to ensure you’re well prepared for the changing way of work.”
Use in Compliance with Existing Laws and Policies for Data Protection & Security. Designated teams or individuals should revisit internal policies and procedures to ensure that they account for planned or permitted uses of generative AI. Employees must understand that relevant current or pending legal obligations apply to the use of new tools.
Employee Training and Education. Identified personnel should inform employees of the implications and consequences of using generative AI tools in the workplace, including providing training and resources on responsible use, risk, ethics, and bias. Designated leads should provide employees with regular reminders of legal, regulatory, and ethical obligations.
Employee Use Disclosure. Organizations should provide employees with clear guidance on when and whether to use organizational accounts for generative AI tools, as well as policies regarding permitted and prohibited uses of those tools in the workplace. Designated leads should communicate norms around documenting use and disclosing when generative AI tools are used.
Outputs of Generative AI. Systems should be implemented to remind employees to verify outputs of generative AI, including for issues regarding accuracy, timeliness, bias, or possible infringement of intellectual property rights. Organizations should determine whether and to what extent compensation should be provided to those whose intellectual property is implicated by generative AI outputs. When generative AI is used for coding, appropriate personnel should check and validate outputs for security vulnerabilities.
Old Laws & New Tech: As Courts Wrestle with Tough Questions under US Biometric Laws, Immersive Tech Raises New Challenges
Extended reality (XR) technologies often rely on users’ body-based data, particularly information about their eyes, hands, and body position, to create realistic, interactive experiences. However, data derived from individuals’ bodies can pose serious privacy and data protection risks for people. It can also create substantial liability risks for organizations, given the growing volume of lawsuits under the Illinois Biometric Information Privacy Act (BIPA) and scrutiny of biometric data practices by the Federal Trade Commission (“FTC” or “Commission”) in their recent Policy Statement. At the same time, there is considerable debate and lack of consensus about what counts as biometric data under existing state privacy laws, creating significant uncertainty for regulators, individuals, and organizations developing XR services.
This blog post explores the intersection of US biometric data privacy laws and XR technologies, particularly whether and to what extent specific body-based data XR devices collect and use may be considered “biometric” under various data protection regimes. We observe that:
Face templates and iris scans used to authenticate an individual’s identity are regulated biometrics, therefore those use cases in XR are covered by biometric laws.
Laws with broad definitions of biometrics may apply to systems that use face detection, as seen in emerging case law from Illinois regarding virtual try-on XR applications.
Organizations have taken steps that reduce their liability risk regarding face-based biometric systems, including by minimizing collection of identifying data or processing biometric data on individuals’ devices.
Other body-based data not used for identification in XR, like eye-tracking and voice analysis, may also be considered “biometric” if the technology and data are capable of identifying an individual.
A. Face Templates, Hand Scans, and Iris Scans Used to Authenticate an Individual’s Identity Are Regulated Biometrics, Therefore User Authentication in XR is Covered by Biometric and Comprehensive Privacy Laws
With the exception of CUBI (and to a certain extent, BIPA), most biometric and comprehensive data privacy statutes tie their definitions of “biometric data” to identification, meaning the laws are intended to regulate unique physiological characteristics that entities use to identify an individual. Generally, each biometric and comprehensive law focuses on five forms of biometric data: retina or iris scan, fingerprint, voiceprint, hand scan, and face scan. BIPA, in particular, applies to “biometric identifiers,” defined as a “retina or iris scan, fingerprint, voiceprint, or scan of hand or face geometry,” as well as “biometric information,” which includes “any information…based on an individual’s biometric identifier used to identify an individual.” Therefore, any entity that uses technology to scan an individual’s iris, finger, hand, or face to uniquely identify an individual (1:many) or authenticate their identity (1:1) must comply with BIPA’s requirements, unless they fall within one of BIPA’s exemptions or exclusions. The same conclusion applies to CUBI, BPPA, and comprehensive data privacy laws.
XR devices often use iris, face, or hand scans to authenticate a user’s identity to log in to their profile or enable in-app payments. Much like computers or smartphones, more than one user may use a single XR device, so authenticating the specific person using the device at a given time allows for more personalization and secure transactions. As a result, iris or face authentication systems in XR devices are likely covered by U.S. biometric and comprehensive data privacy laws. The laws typically require organizations to obtain user consent before enrolling individuals in this sort of biometric XR system, and BIPA has potentially thorny provisions requiring “written consent,” which can be challenging to implement for many XR applications. The face and eye scans that XR technologies use for authentication may also be considered “sensitive data” under comprehensive data privacy laws, such as the California Privacy Rights Act (CPRA) or the Connecticut Data Privacy Act, requiring organizations to provide individuals opt-out rights, including the right to opt out of data sales and other transfers.
Most XR authentication technologies employ live capture of biometrics. Iris, face, or hand scans are captured in real-time when an individual first enrolls, and subsequent scans are likewise captured in real-time when the individual authenticates their identity to the device or app. These scenarios are typically covered by biometrics laws as described above. However, there is some uncertainty regarding biometric laws’ application to XR devices that create a biometric template from non-real-time photos, videos, and audio recordings. Most biometric and comprehensive privacy laws exclude photos, videos, and audio recordings from the scope of “biometric data” to varying degrees (with the exception of CUBI and the CPRA). Utah and Virginia’s comprehensive privacy laws, for example, broadly exempt photographs “or any data generated therefrom” from coverage, making their biometric regulations perhaps less likely to apply to photographic scanning. But case law under BIPA shows that these provisions may not exclude “biometric templates” derived from non-real-time photos, videos, or audio recordings. In Shutterfly v Monroe, the United States District Court for the Northern District of Illinois concluded that narrowly reading “biometric identifier” only to mean real-time scans would swallow the intent of the law, thus photographic scanning to create “templates” were still within scope. Laws like the Connecticut Data Privacy Act (CTDPA) and the final rules for the Colorado Privacy Act (CPA) that do not exclude photos or data generated from these sources if an entity uses them for identification purposes, or CUBI, which contains no exemptions for photographs at all, are likely to follow this analysis. The FTC’s conception of biometric information similarly and explicitly encompasses photos, videos, audio recordings, and certain data derived from these sources, making it likely that most regulators will still consider “biometric templates” created from photographic scanning subject to applicable biometric regulations.
B. Laws with Broad Definitions of Biometrics May Apply to Systems that Use Face Detection, as Seen in Emerging Case Law from Illinois Regarding Virtual Try-On XR Applications
Despite most laws’ goal to regulate biometric data that is uniquely identifying, several statutes’ text can be interpreted to apply to biometric technologies that merely distinguish a face from other objects or analyze facial characteristics, without identifying a particular individual. Depending on a privacy law’s definition of “biometric data,” courts may hold that the term regulates technologies that utilize data derived from an individual’s face, eyes, or voice even when they are not used for identification purposes. In XR, devices may use inward-facing cameras to conduct facial analysis for non-identification purposes, such as rendering expressive avatars. Augmented reality (AR) products like “virtual try-on” may also use facial analysis for people to visualize how different products – like eyeglasses – might look on them. Like many other XR applications, VTO primarily uses facial scans to detect and correctly align the product with an individual’s physical features, rather than for identification purposes.
Some laws with broad definitions can apply to these non-identification technologies unless a specific exception applies. CUBI does not require “biometric identifiers” to uniquely identify an individual, which has prompted the Texas Attorney General to claim that CUBI applies to the capture of face geometry regardless of whether an entity uses these facial maps for individual identification. The FTC’s conception of biometric technologies also broadly encompasses “all technologies that use or purport to use biometric information for any purpose.” But most notably, BIPA is complex because its definition of “biometric identifiers” does not explicitly require that the data be used for identification (in contrast to the statute’s definition of “biometric information,” which does require identification). As a result, Illinois courts have largely found that any facial scan may create a “biometric identifier,” such as with doorbell cameras, photo grouping, and Snapchat filters. This is true even when that technology’s facial scan feature was not used to identify the individual in the photo or video frame.
Recent BIPA lawsuits brought against companies that offer (VTO) illustrate how broad biometric laws might apply to XR devices that use facial analysis. In Theriot v. Louis Vuitton North America, Inc., a federal court permitted BIPA claims to proceed against Louis Vuitton’s VTO sunglasses application, finding that the technology’s use of facial scans was analogous to BIPA case law holding that face scans derived from photographs constitute biometric identifiers. Other VTO cases have had similar outcomes. Only VTO technology used for healthcare-related purposes, such as trying on prescription eyeglasses, have been found by courts to be outside the scope of BIPA. But this result did not rest on BIPA’s overall definition of biometric data, but rather arose from a narrow exception for “information captured from a patient in a health care setting.” So BIPA may not apply to medical providers’ use of XR apps or other immersive technologies, such as brain computer interfaces (BCIs), for diagnostic purposes, but BIPA’s coverage of non-identifying, non-medical uses remains a source of substantial confusion. This confusion undermines individuals’ understanding of their privacy rights and presents liability risks for organizations.
C. Organizations May Reduce their Liability Risk by Minimizing Collection of Identifying Data or Processing Biometric Data on Individuals’ Devices
Some organizations have taken steps to limit their liability risks by minimizing the collection of identifying data or processing biometric data on individuals’ devices. Case law suggests that some facial detection technologies fall outside the scope of BIPA and other biometric regulations if (1) there is no mechanism for the technology to retain facial scans or link scans to a user’s individual identity or account; and/or (2) all of the data is stored on-device.
First, in Daichendt and Odell v. CVS Pharmacy, the Northern District of Illinois dismissed a case against CVS for its passport photosystem, which scans facial geometry in photos to confirm that they meet government requirements for passports (e.g., a person’s eyes are open, their mouth is closed and not smiling, and eyeglasses are not present). The court held that the plaintiffs failed to allege that CVS’ photosystem enabled CVS to determine their identities, nor did the plaintiffs provide CVS “with any information, such as their names or physical or email addresses, that could connect the voluntary scans of face geometry with their identities.”
Separately, in Apple v. Barnett, the Illinois’ appellate court held that Apple was not subject to BIPA requirements regarding their Face ID on iPhone because the company was not “collecting” or “possessing” users’ biometric data since the data was completely stored on the device and never stored on Apple servers. Thus, XR devices that do not retain facial scans that can link to users’ accounts, or only store data on-device (such as Apple’s recently announced Vision Pro) may be out of scope of even some of the broadest biometrics laws.
D. Eye-tracking and Voice Analysis May Also be Considered “Biometric” if the Technology and Data are Capable of Identifying an Individual
In addition to face-based biometric technologies, most XR devices also use other forms of body-based detection or characterization systems for device functionality, such as voice analysis and eye-tracking. As seen with facial detection, these features are developed to detect or create predictions regarding bodily characteristics or behavior, but the subject is typically not identifiable and PII is typically not retained. For example, XR devices often contain microphones to capture a user’s voice and surroundings, which can enable voice commands, verbal interactions with other users, spatial mapping, and realistic sound effects. XR devices may also maintain inward-facing cameras that collect data about a user’s gaze—where they look and for how long—to enable eye tracking. This may be used to improve graphics and allow for more expressive avatars, including avatars that can display microexpressions.
Whether these systems that collect voice or gaze data are covered by biometric or comprehensive data privacy laws may depend on whether an organization can use the technology to identify an individual, even if not used in that capacity. As seen in CVS Pharmacy, many Illinois courts focus on the capacity of the technology to identify an individual. As an initial matter, biometric and comprehensive privacy laws typically apply to “voiceprints,” and not voice recordings. As stated by the Illinois Attorney General, “a voiceprint, which is a record of mechanical measurement, is not the same as a simple recording of a voice.”
However, the line between a voice recording and a voiceprint is blurry, particularly as it relates to the gray area of natural language processing (NLP)—a kind of artificial intelligence (AI) that can use audio to understand, interpret, and manipulate language. In Carpenter v. McDonald’s Corp., the U.S. District Court for the Northern District of Illinois found that McDonald’s drive-through voice assistant technology could be used for identification purposes, and thus could be considered a “voiceprint” under BIPA, since the technology’s patent application states that the technology may capture voice characteristics “like accent, speech pattern, gender, or age for the purpose of training the AI.” In a similar ongoing case against Petco, an Illinois federal judge permitted BIPA claims to proceed regarding employee voice data, stating “[w]hat matters [at the dismissal stage] is not how defendant’s software actually used plaintiffs’ data, but whether the data that Petco gathered was capable of identifying [the employees].” As a result, if an XR device captures vocal characteristics that are capable of unique identification, certain voice data may be considered a “voiceprint” under BIPA. This analytical framework will likely apply to jurisdictions that define biometric data to include biological characteristics that have the potential to identify an individual, such as in the final rules under the Colorado Privacy Act regarding biometric identifiers, or under the FTC’s Policy Statement on Biometric Information.
Whether privacy laws apply to gaze data, however, is even less clear. BIPA lawsuits against online exam proctoring services, autonomous vehicles, and “smart advertising screens” suggest that eye-tracking could be a biometric identifier under BIPA, even if not used for identification. In each of these cases, the technology conducted eye-tracking to determine where a user was looking—whether on the screen, the road, or in the store—but did not identify the individual. Instead, these technologies made inferences about whether someone may be cheating, not paying attention to the road, or what product they were looking at. Plaintiffs in these cases argue that eye-tracking is part of the technology’s collection and analysis of facial geometry, thus making it a “biometric identifier” under BIPA.
Unfortunately, state and federal courts in Illinois have not analyzed whether and to what extent eye tracking, without additional face analysis, constitutes a biometric identifier, nor whether it is a subset of facial analysis. Rather, most cases proceed based solely on the software’s overall facial analysis features, if at all. If courts are prone to equate facial detection scans to “facial geometry,” and voice analysis to “voiceprints,” they may also conflate eye tracking with “a retina or iris scan,” and thus treat eye tracking as a biometric identifier. Or they may follow the BIPA plaintiffs’ analysis, lumping eye-tracking into facial analysis as “facial geometry.” Alternatively, courts could characterize eye tracking as altogether separate from BIPA’s “facial geometry” and “retina or iris scan” categories. In any event, like with voice analysis, if an XR device collects gaze data that could be used for identification purposes, laws with broad biometrics definitions will apply, while other laws that have narrower definitions focused on the data or technology’s current use, may exclude the technology.
Takeaways
Statutory language and court opinions vary in how they define and/or apply to biometric data and identifiers. Though the plain text of most U.S. biometric and comprehensive data privacy laws tie their definition of a “biometric” to the identification of an individual, some laws may be more broadly applied to technologies that use body-based data for non-identification purposes. While most of the body-based data XR collects is not used for identification, litigation brought under BIPA and other state laws suggest that lawmakers and judges may consider certain kinds and uses of such data—for example, AR “facial scans,” eye tracking, and voice— to be biometrics. Whether this will be the case (or continue to be the case) depends on how policymakers draft these laws, and how courts, enforcement bodies, and other parties to litigation interpret statutes regulating biometrics.
Insights into Brazil’s AI Bill and its Interaction with Data Protection Law: Key Takeaways from the ANPD’s Webinar
Authors: Júlia Mendonça and Mariana Rielli
The following is a guest post to the FPF blog by Júlia Mendonça, Researcher at Data Privacy Brasil, and Mariana Rielli, Institutional Development Coordinator at Data Privacy Brasil. The guest blog reflects the opinion of the authors only. Guest blog posts do not necessarily reflect the views of FPF.
On July 6, 2023, the Brazilian National Data Protection Authority (ANPD) held a webinar event entitled: The interplay between AI regulation and data protection. The dialogue unfolded in the broader context of developments in AI regulation in Brazil which has, as its main drivers, the bills that propose a Regulatory Framework for Artificial Intelligence in the country. The bills were jointly analyzed by a Commission of 18 jurists appointed by the Federal Senate, which promoted meetings, seminars, and public hearings to substitute them with a new draft proposal. At the beginning of May, the draft produced by the Commission was transformed into a new bill that is currently going through the legislative process: Bill PL nº2338 (AI draft bill).
The ANPD, noting the need to harmonize any upcoming AI regulation with the existing data protection regime (as well as future enforcement matters), organized this webinar, in addition to having published a preliminary analysis of the AI draft bill. The discussions during the webinar offer a glimpse into the AI lawmaking and policymaking in Brazil, one of the largest jurisdictions in the world – one that is also covered by a general data protection law applicable to personal data processed in the context of an AI system. This brief blog post outlines the main topics discussed during the event, particularly in relation to the interplay between the current AI draft bill and Brazil’s General Data Protection Law (LGPD).
The webinar’s opening welcomed Waldemar Gonçalves (President, ANPD, Brazil), Eduardo Gomes (Senator of the Republic, Brazil), and Estela Aranha, (Special Advisor, Ministry of Justice and Public Security, Brazil). The panel that followed was formed by representatives of the National Data Protection Council (CNPD) – a multisectoral advisory body, part of the ANPD structure – namely, Ana Paula Bialer (Founding Partner, Bialer Falsetti Associados, Brazil), Bruno Bioni, (Director and Founder, Data Privacy Brasil), Fabrício da Mota (Vice President, Conselho Federal da OAB, Brazil), and Laura Schertel (Visiting researcher, Goethe Universität Frankfurt; and private law Professor and lawyer, Brazil/EU).
Key representatives highlight the need for ongoing harmonization between AI regulation and data protection law in Brazil
As the President of the ANPD, Waldemar Gonçalves highlighted the Authority’s ongoing work on the AI agenda, noting that data protection rules under the LGPD are closely interconnected with those provided for in the AI draft bill, such as with regard to the right to information. With such similarities in mind, Gonçalves noted the need for harmonization between different tools, such as the Data Protection Impact Assessment (DPIA) and the Algorithm Impact Assessment (AIA).
Another initiative of the ANPD highlighted by Gonçalves as relevant to the AI agenda and the current AI regulatory efforts was the technical agreement between the Authority and the Latin American Development Bank (CAF), which will include a regulatory sandbox pilot program on data protection and AI.
ANPD’s current president closed his remarks recalling the various recent cases in which data protection authorities around the world have spoken out on issues concerning AI-based systems, thereby reinforcing the importance of the ANPD in assuming an active role in this discussion. Eduardo Gomes, rapporteur of the AI draft bill, started from the same premises to support the efforts with the president of the Senate, Rodrigo Pacheco. In addition to reinforcing the importance of work of the Commission of Jurists in laying the groundwork for the debate in Brazil, he also recognized the need to foster other opportunities to “mature the subject.”
Concluding the opening panel, Estela Aranha focused her presentation on the topic of algorithmic discrimination in the context of the interplay between AI and existing data protection norms. Aranha mentioned examples with regards to data mining and how the resulting massive collection of data can generate the most varied risks, including risks of discrimination, and can go beyond the most obvious examples of sensitive and inferred data. The relevance of this specific point in the debate stems from the fact that the proposed AI draft bill is quite detailed, both in terms of definitions and obligations created, with regards to direct and indirect discrimination potentially created or enhanced by AI systems in the Brazilian context. Finally, Aranha also reaffirmed the Ministry of Justice’s support for the Bill.
A deeper dive into the proposed AI draft bill and possible future(s) of AI regulation
The following panel focused on a deeper look at the proposed AI draft bill and some of the specific provisions therein. The first panelist, Ana Paula Bialer, highlighted that there is already a robust framework for data protection that grants the data subject greater control over their data, based on the principle of “informational self-determination.” However, Bialer made a point that there may be a certain difficulty in applying the rationale of data protection to AI. Not in the sense that the data used is presumably not protected, but rather that there should be a thorough exercise of extension and “revalorization” of the principles of the LGPD, combined with a review of the set of rights put in place in the context of AI systems.
Already assessing the current draft bill, Bialer also considered that the meaning of a human-centered approach can be different when thinking about different applications of AI in varying socio-economic contexts, exemplifying her reflection through the topic of recruitment and new hires’ selection and the right to full employment in Brazil. Bialer concluded by reaffirming the benefits that can be brought by AI for social and economic development in the country, as well as for the exercise of fundamental rights. In this context, Bialer welcomed the ANPD’s regulatory sandbox initiative and positioned herself more favorably to a strongly risk-based approach to AI regulation.
Bruno Bioni began by emphasizing the importance of having a dose of skepticism with regards to the broader debate – both on AI, and in respect to AI regulation – especially in a scenario where the almost “apocalyptic” narrative around AI continues gaining notoriety. This is important because, in Bioni’s opinion, such discourse may end up underestimating the regulatory tools that already exist. The very field of personal data protection has already provided positive and negative lessons when it comes to an object of regulation that is very plastic and polyvalent, “with a regulatory mission that is transversal and not sectoral.”
Bioni continued by pointing out that the intersection of data protection, AI regulation, and governance is very much related to the idea of a “toolbox” that opens opportunities for a more collaborative, collective regulatory production, relying on companies themselves to participate and to some extent, be rewarded, for example, if they demonstrate a good level of accountability.
Among the various existing tools and how they can support each other, Bioni highlighted Algorithmic Impact Assessments (AIA) and Data Protection Impact Assessments (DPIAs) as documentation that can foster and unfold into the other in such a way as to optimize both. The ANPD has already positioned the DPIA prescribed by the LGPD as an instrument to be better regulated and better standardized, which, for the expert, will be a significant advancement, even in a hypothetical scenario where it takes a long time for an AI regulation to be passed.
According to Bioni, it is for this reason that data protection authorities around the world have led enforcement actions, in the absence of AI laws or authorities created with this specific mission. Bioni concluded his remarks by pointing out that it is essential to think about a more collective or networked governance approach.
Fabrício da Mota Alves focused on the issue of institutional arrangements and of thinking about future legislation inserted in a regulatory environment that is founded on the administrative action of the Brazilian State. Fabrício pondered on the possibility that, following other countries in the world, the ANPD promotes some degree of administrative action (supervisory and sanctioning, in addition to regulation and awareness) related to AI, reinforcing that there is a concern to understand and call for the ANPD to build a very robust regulatory environment. Above all, there is a call for formal protocols so that companies and experts can understand the limits and the scope of ANPD’s actions in this dynamic scenario.
Celebrating the space provided by the webinar as one of the first and most qualified to take place outside of the legislative environment, Alves emphasized that it is imperative that, also in the context of regulating and enforcing AI-related cases (regardless of specific frameworks), the Brazilian ANPD maintains the stance it has adopted so far, with broad public participation, hearings, public consultations, and processes that are open to criticism from all affected sectors.
What’s next for the Brazilian AI bill?
Brazil’s AI draft bill is in its early stages, although it has already been the result of lengthy discussions by the expert committee assigned to prepare a new draft in 2022. There is an expectation that it will now be analyzed by a special committee of parliamentarians designated specifically to debate the Bill, with the prospect of new rounds of public hearings. After the text is approved by the plenary of the Brazilian Senate, the proposal still goes through the Chamber of Deputies, the reviewing house, until a common text is reached, which will then be sanctioned by the President of the Republic.
The whole webinar, in Portuguese, can be watched here.
Newly Updated Report: The Spectrum of Artificial Intelligence – Companion to the FPF AI Infographic
Artificial Intelligence (AI) has become an integral part of our lives, transforming how we interact, work, and make decisions. From virtual assistants and recommendation systems to autonomous vehicles and medical diagnostics, AI technologies have made remarkable progress. However, as AI continues to advance, it is essential to understand how it works and the ethical considerations that accompany it.
This updated report places generative AI within the larger AI Landscape to address foundational questions about the operation and development of the technology, including generative AI’s use of personal information, the ability of individuals to meaningfully utilize access, correction, or deletion rights, as well as means and methods available to minimize inaccurate information and hallucinations in outputs.
As generative AI becomes mainstream through tools such as Open AI’s ChatGPT and Google’s Bard, it introduces new and transformational use cases for AI in everyday life, including the workplace. However, there are also risks and ethical considerations to manage throughout the lifecycle of these systems. A better understanding of all the kinds of AI systems and how they relate to one another benefits organizations, policymakers, and the general public is essential. The re-release of The Spectrum of Artificial Intelligence – Companion to the FPF AI Infographic strives to do this.
The First Japan Privacy Symposium: G7 DPAs discussed their approach to reign in AI, and other regulatory priorities
The Future of Privacy Forum and S&K Brussels LPC hosted the first Japan Privacy Symposium in Tokyo, on June 22, 2023, following the G7 Data Protection and Privacy Commissioners roundtable. The Symposium brought global thought leadership on the interaction of data protection and privacy law with AI, as well as insights into the current regulatory priorities of the G7 Data Protection Authorities (DPAs) to an audience of more than 250 in-house privacy leaders, lawyers, consultants and journalists from Japan and the region.
The program started with a keynote address from Commissioner Shuhei Ohshima (Japan’s Personal Information Protection Commission), who shared details about the results of the G7 DPAs Roundtable from the day before. Two panels followed, featuring Rebecca Kelly Slaughter (Commissioner, U.S. Federal Trade Commission), Wojciech Wiewiórowski (European Data Protection Supervisor, EU), Philippe Dufresne (Federal Privacy Commissioner, Canada), Ginevra Cerrina Feroni (Vice President of the Garante, Italy), John Edwards (Information Commissioner, UK), and Bertrand du Marais (Commissioner, CNIL, France). Jules Polonetsky, FPF CEO, and Takeshige Sugimoto, Managing Partner at S&K Brussels LPC and FPF Senior Fellow, hosted the Symposium.
The G7 DPA Agenda, built on three pillars: Data Free Flow with Trust, emerging technologies, and enforcement cooperation
The DPAs of the G7 nations started to meet annually in 2020, following the initiative of the UK’s Information Commissioner Office during UK’s G7 Presidency that year. This is a new venue for international cooperation of DPAs, limited to Commissioners from Canada, France, Germany, Italy, Japan, the United Kingdom, the United States, and the European Union. Throughout the year, the DPAs maintain a permanent channel of communication and implement a work plan adopted during their annual Roundtable.
In his keynote at the Japan Privacy Symposium, Commissioner Shuhei Oshshima laid out the results of this year’s Roundtable, held in Tokyo on June 20 and 21. The Commissioner highlighted three pillars guiding the group’s cooperation this year: (I) Data Free Flow with Trust (DFFT), (II) emerging technologies, and (III) enforcement cooperation.
The G7 Commissioners’ Communique expressed overall support for the DFFT political initiative, welcoming the reference to DPAs as stakeholders in the future Institutional Arrangement for Partnership (IAP), a new structure the G7 Digital Ministers announced earlier in April to operationalize the DFFT. However, in the Communique, the G7 DPAs emphasized that they “must have a key role in contributing on topics that are within their competence in this Arrangement.” It is noteworthy that, among their competencies, most G7 DPAs have the authority to order the cessation of data transfers across borders if legal requirements are not met (see, for instance, this case from the CNIL – the French DPA, this case from the European Data Protection Supervisor, or this case from the Italian Garante).
The IAP seems to provide a key role for governments themselves currently, in addition to stakeholders and “the broader multidisciplinary community of data governance experts from different backgrounds,” according to Annex I of the Ministerial Declaration announcing the Partnership. The DPAs are singled out only as an example of such experts.
In the Action Plan adopted in Tokyo, the G7 DPAs included clues as to how they see the operationalization of DFFT playing out: through interoperability and convergence of existing transfer tools. As such, they endeavor to “share knowledge on tools for secure and trustworthy transfers, notably through the comparison of Global Cross-Border Privacy Rules (CBPR) and EU certification requirements, and through the comparison of existing model contractual clauses.” (In an analysis touching broadly beyond the G7 jurisdictions, the Future of Privacy Forum published a report earlier this year emphasizing many commonalities, but also some divergence, among three sets of model contractual clauses proposed by the EU, the Iberoamerican Network of DPAs, and ASEAN).
Arguably, though, DFFT was not the main point on the G7 DPAs agenda. They had adopted a separate and detailed Statement on generative AI. In his keynote, Commissioner Shuhei Ohshima remarked that “generative AI adoption has increased significantly.” In order to promote trustworthy deployment and use of the new technology “the importance of DPAs is increasing also on a daily basis,” the Commissioner added.
Generative AI is not being deployed in a legislative void, and data protection law is the immediately applicable legal framework
Top of mind for G7 data protection and privacy regulators is AI, and generative AI in particular. “AI is not a law-free zone,” said FTC Commissioner Slaughter during her panel at the Symposium, being very clear that “existing laws on the books in the US and other jurisdictions apply to AI, just like they apply to adtech, [and] social media.” This is apparent across the G7 jurisdictions: in March, the Italian DPA issued an order against OpenAI to stop processing personal data of users in Italy following concerns that ChatGPT breached the General Data Protection Regulation (GDPR); in May, the Canadian Federal Privacy Commissioner opened an investigation into ChatGPT jointly with provincial privacy authorities; and, in June, Japan’s PIPC issued an administrative letter warning OpenAI that it needs to comply with requirements from the Act on the Protection of Personal Information, particularly regarding the processing of sensitive data.
At the Japan Privacy Symposium, Ginevra Cerrina Feroni, VP of the Garante, shared the key concerns guiding the agency’s enforcement action against OpenAI, which was the first such action in the world. She highlighted several risks, including a lack of transparency about how OpenAI collects and processes personal data to deliver the ChatGPT service; uncertainty regarding a lawful ground for processing personal data, as required by the GDPR; a lack of avenues to comply with the rights of data subjects, such as access, erasure, and correction; and, finally, the potential exposure of minors to inappropriate content, due to inadequate age gating.
After engaging in a constructive dialogue with OpenAI the Garante suspended the order, seeing improvements in previously flagged aspects. “OpenAI published a privacy notice to users worldwide to inform them how personal data is used in algorithmic training, and emphasized the right to object to such processing,” the Garante Vice President explained. She continued, noting that OpenAI “provided users with the right to reject their personal data being used for training the algorithms while using the service, in a dedicated way that is more easily accessible. They also enabled the ability of users to request deletion of inaccurate information, because – and this is important – they say they are technically unable to correct errors.” However, Vice President Cerrina Feroni mentioned that the investigation is ongoing and that the European Data Protection Board is currently coordinating actions among EU DPAs on this matter.
The EDPS added that purpose limitation is among his chief concerns with services like ChatGPT, and generative AI more broadly. “Generative AI is meant to advance communication with human beings, but it does not provide fact-finding or fact-checking. We should not expect this as a top feature of Large Language Models. These programs are not an encyclopedia; they are just meant to be fluent, hence the rise of possibilities for them to hallucinate,” Supervisor Wiewiorowski said.
Canadian Privacy Commissioner Philippe Dufresne emphasized that how we relate to generative AI from a privacy regulatory perspective “is an international issue.” Commissioner Dufresne also added, “a point worth repeating is that privacy must be treated as a fundamental right.” This is important, as “when we talk about privacy as a fundamental right, we point out how privacy is essential to other fundamental human rights within a democracy, like freedom of expression and all other rights. If we look at privacy like that, we must see that by protecting privacy, we are protecting all these other rights. Insofar as AI touches on these, I do see privacy being at the core of all of it,” Commissioner Dufresne concluded.
The G7 DPAs’ Statement on Generative AI outlines their key concerns, such as lack of legal authority to process personal data at all stages
In the aforementioned Generative AI Statement, the G7 data protection regulators laid out their main concerns in relation to how personal data is processed through this emerging type of computer program and service. First and foremost, the commissioners are concerned that processing of personal data lacks legal authority during all three relevant stages of developing and deploying generative AI systems: for the data sets used to train, validate and test generative AI models; for processing personal data resulting from the interactions of individuals with generative AI tools during their use; and, for the content that is generated by generative AI tools.
The commissioners also highlighted the need for security safeguards to protect against threats and attacks that seek to invert generative AI models, and that would technically prevent extractions or reproductions of personal data originally processed in datasets used to train the models. They also advocated for mitigation and monitoring measures to ensure personal data created by generative AI is accurate, complete, and up-to-date, as well as free from discriminatory, unlawful, or otherwise unjustifiable effects.
It is clear that data protection and privacy commissioners are proactive about ensuring generative AI systems are compatible with privacy and data protection laws. Only two weeks after their roundtable in Tokyo, it was reported that the US FTC initiated an investigation against OpenAI. And this proactive approach is intentional. As UK’s Information Commissioner, John Edwards, made clear, the commissioners are “keen to ensure” that they “do not miss this essential moment in the development of this new technology in a way that [they] missed the moment of building the business models underpinning social media and online advertising.” “We are here and watching,” he said.
Regardless of the adoption of new AI-focused laws, DPAs would remain central to AI governance
The Commissioners also discussed the wave of legislative initiatives targeting AI in their jurisdictions. AI systems are not built and deployed in a legislative void: data protection law is largely and immediately relevant, as is consumer protection law, product liability rules, and intellectual property law. In this environment, what is the added value of specific, targeted legislation addressing AI?
Addressing the EU AI Act proposal, European Data Protection Supervisor Wiewiórowski noted that the EU’s initiation of the legislation is not because the legislator thought there was a vacuum. “We saw that there were topics to be addressed more specifically for AI systems. There was a question whether we approach it as a product, service, or some kind of new phenomenon as far as legislation is concerned,” he added. As for the role of the DPAs once the AI Act will be adopted, he brought up the fact that in the EU, data protection is a fundamental right: which means that all legislation or policy solutions governing processing of personal data in a way or another must be looked at through this lens. As supervisory authorities tasked with guaranteeing this fundamental right, DPAs will continue playing a role.
The framework ensuring the enforcement of the AI Act is still under debate, as EU Member States are tasked with designating competent national authorities, and the European Parliament hopes to create a supranational collaborative body to play a role in enforcement. However, one thing is certain: in the proposal, the EDPS has been designated the competent authority to ensure that EU agencies and bodies comply with the EU AI Act.
The CNIL seems to be eyeing the designation as EU AI Act enforcer as well. Commissioner du Marais pointed out that “since 1978, the French Act on IT and Freedom has banned automated decisions. We have a fairly long and established body of case law.” Earlier this year, the CNIL created a dedicated department including data and computer scientists among staff to monitor how AI systems comply with legal obligations stemming from data protection law. “To be frank, we don’t know yet what will come out of the legislative process, but we have started to prepare ourselves. We have also been designated by domestic law as supervisory and certification authority for AI during the 2024 Olympic Games.”
The Garante has a long track record of enforcing data protection law on algorithmic systems and decision-making that impacted the rights of individuals. “The role of the Garante in safeguarding digital rights has always been prominent, even when the issue was not yet widely recognized by the public,” said Vice President Cerrina Feroni. Indeed, as shown by extensive research published last year by the Future of Privacy Forum, European DPAs have long been enforcing data protection law in cases where automated decision-making was central. The Garante led impactful investigations against several gig economy apps and their algorithms’ impacts on people.
Canada is also in the midst of legislating AI, introducing a bill last year that is currently under debate. “There is similarity with the European proposal, but [the Canadian bill] focuses more on high impact AI systems and on preventing harms and biased outputs and decision-making. It provides significant financial fines,” Commissioner Dufresne explained. As part of the bill, enforcement is currently assigned to the relevant ministry in the Canadian government. The Privacy Commissioner explained that the regulatory activity would be coordinated with his office, but also with the competition, media, and human rights regulators in Canada. When contributing recommendations during the legislative process, Commissioner Dufresne noted that he suggested “privacy to be a key principle.” In light of his vision that privacy as a fundamental right is essential for the realization of other fundamental rights, the Commissioner had a clear message that “the DPAs need to be front and center” of the future of AI governance.
UK Commissioner Edwards echoed the value of entrenched collaboration among digital regulators, adding that the UK already has an official “Digital Regulators Cooperation Forum,” established with its own staff. The entity “is important to provide a coherent regulatory framework,” he said.
Children’s privacy is a top priority across borders, with new regulatory approaches showing promising results
One of the key concerns that the G7 DPAs have in relation to generative AI is how the new services are dealing with children’s privacy. In fact, the regulators have made it one of their top priorities to broadly pursue the protection of children’s privacy when regulating social media services, targeted advertising, or online gaming, among others.
Building on a series of recent high-profile cases brought by the FTC in this space, Commissioner Slaughter couldn’t have been clearer: “Kids are a huge priority issue for the FTC.” She reminded the audience that COPPA (Children’s Online Privacy Protection Act) has been around for more than two decades, and it is one of the strongest federal privacy laws in the US: “The FTC is committed to enforcing it aggressively.” Commissioner Slaughter explained that the FTC’s actions, such as their recent case against Epic Games, include considerations related to teenagers as well, even if they are not technically covered by COPPA protections, but are covered by the “unfair practices” doctrine of the FTC.
UK Commissioner John Edwards gave a detailed account of the impact of the UK’s Age Appropriate Design Code in the design of online services provided to children, which was launched by his office in 2020. “We have seen genuine changes, including privacy settings being automatically set to very high for children. We have seen children and parents and carers being given more control over privacy settings. And we have seen that children are no longer nudged to lower privacy settings, with clearer tools and steps in place for them to exercise their data protection rights. We have also seen ads blocked for children,” Commissioner Edwards said, pointing out that these are significant improvements for the online experience of children. These results have been obtained primarily through a collaborative approach with the service providers, who have implemented changes after their services were subject to audits conducted by the regulator.
Children’s and teenagers’ privacy is also top of mind for the CNIL. Among a series of guidance, recommendations, and actions, the French regulator is adding another layer to its approach – digital education. “We have made education a strategic priority. We have a partnership with the Ministry of Education and we have available a platform to certify digital skills for children, as well as with resources for kids and parents,” Commissioner du Marais said. Regarding regulatory priorities, he emphasized attention to age verification tools. Among the principles the French regulator favors for age verification are no direct collection of identity documents, no age estimates based on web browsing history, and no processing of biometric data to recognize an individual. The CNIL has asked websites not to carry out age verification themselves, and to instead rely on third-party solutions.
The discussions of the G7 DPA Commissioners who participated in the first edition of the Japan Privacy Symposium laid out a vibrant and complex regulatory landscape, centered around new challenges posed to societal values and rights of individuals by AI technology, but also making advancements in perennial topics like cross-border data transfers and children’s privacy. More meaningful and deeper enforcement cooperation is to be expected among the G7 Commissioners, whose Action Plan espoused their commitment to move towards constant exchanges related to enforcement actions and to revitalize existing global enforcement cooperation networks, like GPEN (Global Privacy Enforcement Network). Next year, the G7 DPA Commissioners will meet in Rome.
Editor: Alexander Thompson
A New Domicile for Comprehensive Privacy in Delaware
On June 30, 2023, in the final hours of the Delaware legislative session, lawmakers in Dover passed House Bill 154, the Delaware Personal Data Privacy Act (“DPDPA”). If enacted by Governor Carey, the DPDPA will take effect on January 1, 2025 and follows the general model established by the Connecticut Data Privacy Act (CTDPA), with some notable differences. Delaware will become the twelfth U.S. state to adopt a comprehensive data privacy law to govern the collection, use, and transfer of personal data.
1. Broad Scope
The DPDPA establishes the lowest primary coverage threshold of any state comprehensive privacy law passed so far, applying to organizations that control or process the data of at least 35,000 Delaware residents annually. Typically, state-level comprehensive privacy laws cover organizations that control or process the data of at least 100,000 state citizens each year. The DPDPA’s scope was likely tailored to fit Delaware’s small size and population: by land area Delaware is smaller than any other U.S. state save Rhode Island, and has one of the lowest populations in the country, estimated by U.S. Census data at 1.018 million in 2022.
The Act exempts specific data that is subject to existing laws, including Health Insurance Portability and Accountability Act (HIPAA) and Fair Credit Reporting Act (FCRA)-covered data while broadly carving out Gramm-Leach-Bliley Act (GLBA) covered entities. However, the DPDPA diverges from most other state-level comprehensive privacy laws by not broadly exempting non-profits or higher education institutions.
2. Timely Sensitive Data Categories
The DPDPA establishes a category of “sensitive” personal information that is subject to greater protections, which includes categories such as “[d]ata revealing racial or ethnic origin,” “religious beliefs,” and “[p]recise geolocation data.” However, the DPDPA expands this list beyond that seen in many other states, including “status as transgender or nonbinary,” which is also recognized as a sensitive information category in Oregon’s recently-passed comprehensive privacy law, and “mental or physical health condition or diagnosis (including pregnancy).”
Although all currently enacted comprehensive privacy laws recognize some version of “mental or physical health condition or diagnosis” as sensitive, the DPDPA is the first state-level comprehensive privacy law to explicitly include pregnancy as a category of sensitive data. The recently-passed Connecticut Senate Bill 3 (SB 3), which partially updates the Connecticut Data Privacy Act (CTDPA), also specifically classifies data related to pregnancy and reproductive health as sensitive. Both SB 3 and the DPDPA likely reflect lawmaker focus on the privacy of reproductive health and pregnancy data in the wake of the Supreme Court’s overturning of Roe v. Wade.
3. Protections for Teens
The DPDPA forbids covered entities from selling or processing for targeted advertising purposes the data of consumers that the controller knows, or willfully disregards, are between the ages of 13 and 17 without consent. This prohibition goes farther than similarly-structured prohibitions in California, Connecticut, and Montana, which place restrictions on the sale and processing of the data of consumers between the ages of 13 and 15. The DPDPA’s broader coverage of teen’s data reflects the ongoing attention to youth privacy that has permeated state legislatures this session. While this is the first time a state-level comprehensive privacy law has structured it’s protections to cover teens up to the age of 17 (although CT SB 3 creates similar protections for 13-17-year olds), child-directed privacy and online safety laws, including the California Age-Appropriate Design Code and Utah Senate Bill 152, have increasingly applied to the data and activity of teenagers up to age 17.
4. Expanded Rights to Access and Delete
In line with other comprehensive privacy laws, the DPDPA grants consumers the right to require controllers to delete their personal data. Unlike comparable laws, however, the DPDPA requires controllers to delete data obtained about a person from a third-party source (such as a data broker) except for “a record of the deletion request and the minimum data necessary for the purpose of ensuring the consumer’s personal data remains deleted from the controller’s records,” which they may not use for any other purpose. In contrast, other state privacy laws typically permit controllers to retain data obtained about a person from third-party sources so long as they opt that person out of the processing of their personal data for all non-exempt purposes. The DPDPA also creates a unique affirmative right to “obtain a list of the categories of third parties to whom the controller has disclosed the consumer’s personal data.”
5. Unique Treatment of Nonprofits
Delaware joins Colorado and Oregon in not generally carving out nonprofit organizations in its scope. Like Oregon, however, the Delaware law carves out nonprofits that combat insurance fraud. The DPDPA also creates a novel data-level exemption for the “[p]ersonal data of a victim of or witness to child abuse, domestic violence, human trafficking, sexual assault, violent felony, or stalking that is collected, processed, or maintained by a nonprofit organization that provides services to victims of or witnesses to child abuse, domestic violence, human trafficking, sexual assault, violent felony, or stalking.”
6. UOOM Uncertainty
The DPDPA would be the seventh comprehensive state privacy law to permit consumers to exercise certain rights on a default basis through what is commonly known as a “Universal Opt-Out Mechanism” (UOOM), joining California, Colorado, Connecticut, Montana, Texas, and Oregon. The UOOMs that are currently in use often take the form of a browser extension, which sends out an automatic signal to web pages visited by a consumer with the extension enabled, notifying it that they would like to exercise a certain consumer right.
The DPDPA establishes that consumers have the right to opt out of the processing of their personal information for: targeted advertising, data sales, and profiling for the purposes of automated decision-making with significant impact on the consumer. Drafting ambiguities make it unclear whether the DPDPA permits opting-out of profiling via device signals, which would be a first for a state comprehensive privacy law. The DPDPA does not allow for rulemaking.
FPF Paper, “The Thin Red Line …,” Receives the Council of Europe’s 2023 Stefano Rodotà Award
On Friday, June 16th, members of the FPF team joined the 44th Plenary meeting of the Council of Europe’s Committee of Convention 108 in Strasbourg, France to accept a tremendous research honor. On this occasion, Katerina Demetzou, Senior Counsel for Global Privacy, Dr. Gabriela Zanfir-Fortuna, VP for Global Privacy, and Sebastião Barros Vale, former Senior Counsel for Europe at FPF, received the 2023 Stefano Rodotà Data Protection award in the category of ‘best academic article’ for their paper, “The Thin Red Line: Refocusing Data Protection Law on Automated Decision-Making, A Global Perspective with Lessons from Case-Law.” Demetzou and Barros Vale were present in Strasbourg during the Plenary meeting to present the paper and lift the award.
The Council of Europe (CoE), founded in 1949, is an international organization with 46 Member States and 6 Observer States. All Council of Europe Member States have signed up to the European Convention of Human Rights (ECHR), a treaty designed to protect human rights, democracy, and the rule of law. The European Court of Human Rights (ECtHR) oversees the implementation of the ECHR in the Member States.
Demetzou, Barros Vale, and Dr. Gabriela Zanfir-Fortuna, VP for Global Privacy at FPF, are honored to receive recognition at the birthplace of the CoE’s historic 1981 treaty, the Convention for the Protection of Individuals with Regard to Automatic Processing of Personal Data. Convention 108 established the first legally binding international instrument in data protection, and the CoE adopted the modernized Convention 108+ in 2018. A year later, the Convention 108+ Committee established the Stefano Rodotà Award to honor the memory and legacy of Stefano Rodotà (1933-2017), a leading Italian Professor and politician, and one of the founding fathers of data protection law in Europe. The Stefano Rodotà Award is awarded to precedent-setting and innovative research in the field of data protection. The dedicated academic, intellectual, and political influence of President Rodotà lives on through Demetzou, Barros Vale, and Dr. Zanfir-Fortuna’s global exploration of data protection instruments safeguarding individuals against harms from ADM in emerging technologies.
“The Thin Red Line” analyzes the legal protection provided by data protection law to individuals that are being subjected to automated decision making (ADM) on the basis of their personal data. To that end, the authors dedicate the first part of their research to an analysis of European Data Protection Authority (DPAs) enforcement actions of the GDPR on ADM cases. The second section looks to Brazil, Mexico, Argentina, Colombia, China and South Africa to explore protections against harmful ADM found in non-EU general data protection laws. The article concludes that even in cases where a processing operation does not meet the high threshold set by Article 22 GDPR (‘solely by automated means’), DPAs have made use of an array of legal principles, rights, and obligations to protect individuals against ADM practices, nonetheless. With the exception of Colombia, the other non-EU jurisdictions have a specific ADM provision. In all cases studied, the general data protection laws provide a broad material scope, such that any automated processing operation, solely or not, is regulated according to relevant provisions. Additionally, all laws studied include strong transparency and fairness requirements.
While the debate on a European Regulation for AI is ongoing, this paper aims to contribute to the discussion by highlighting that in cases where algorithms and AI systems process personal data, the GDPR is enforceable and protects individuals. Despite extensive legal scholarship on Article 22 GDPR, FPF’s experts identified a gap in previous literature through their global examination of existing enforcements and interpretations from regulators.
After the award ceremony, Demetzou was especially grateful for “the Committee’s warmth, as well as their committed understanding and appreciation for our research.” Dr. Zanfir-Fortuna called on the importance of the article’s findings while reflecting on emerging AI regulatory trends: “Data protection law has proved to be one of the most relevant existing legal frameworks to deal with the risks posed by the mass deployment of new AI tools. Existing legal obligations related to processing of personal data, on all continents, are stringent and more pressing than possible future AI legislation, as they are immediately applicable to existing AI systems.” The authors hope this intervention, as well as the paper’s global scan, will support researchers and policymakers in understanding how existing data protection law protects against potential harms from algorithms and AI systems.
Melis Ulusel is a current student at the George Washington University Law School and an FPF Policy Intern.
A significant new chapter for data privacy protections in the United States will commence on July 1, 2023 as broad-based consumer privacy laws in Colorado and Connecticut take effect. At the same time, the California Privacy Rights Act amendments to the California Consumer Privacy Act of 2018 will become enforceable.
Over the next 3 years, an additional 11 recently enacted significant state privacy laws are scheduled to take effect. To assist stakeholders in tracking the emerging state privacy landscape, the Future of Privacy Forum has prepared a chart listing the primary and secondary effective dates for these statutes.
FPF Files Comments with the U.S. Department of Health and Human Services (HHS) Office for Civil Rights
On June 15, the Future of Privacy Forum (FPF) filed comments with the U.S. Department of Health and Human Services (HHS) Office for Civil Rights (OCR) regarding the Notice of Proposed Rulemaking (NPRM) on extending additional protections to reproductive health care data under the Health Insurance Portability and Accountability Act (HIPAA).
One year ago last week, the Supreme Court issued a decision that has resulted in loss of access to reproductive care for many Americans. Federal and state legislative and regulatory entities have been quick to respond to protect rights to reproductive care, a fundamental aspect of decisional privacy. Rulemakings such as this one by HHS OCR seek to fill the gap left in the wake of the Supreme Court’s 2022 decision that fundamentally shifted the landscape of data and information privacy. With a post-Dobbs lens, FPF has filed comments on this rulemaking based on the following recommendations.
We recommend that HHS bolster privacy safeguards and support the responsible handling of reproductive health care information (RHCI) by specifically:
Ensuring that covered entities are aware of and responsible for information that, directly or indirectly, can reveal data about individuals seeking or receiving reproductive health care;
Providing additional guidance and resources to address the information privacy responsibilities of covered entities for their business associates and vendors;
Distributing privacy education and guidance materials to covered entities and partners on data privacy transparency;
Conducting regulatory analysis and providing compliance support for small clinics and rural/remote providers facing increased legal requests for reproductive and related health information;
Addressing privacy protections for reproductive health care data collected and generated during and as a part of clinical research.
FPF’s full comments to the HHS are available here.
