Alicia Solow‑Niederman is a leading scholar at the forefront of algorithmic accountability, data governance, and information privacy. As an Associate Professor at George Washington University Law School, she examines how emerging technologies, especially artificial intelligence, expose the limits of existing legal frameworks and reveal deeper questions about power, governance, and the values embedded in our regulatory systems. In this interview, Alicia discusses the challenges of governing AI across overlapping legal regimes, the politics of technical standards, and the evolving role of courts, agencies, and private actors in shaping the digital landscape. Her insights illuminate the tensions among privacy, transparency, and innovation, offering a nuanced view of how law can adapt to technological change while remaining grounded in democratic principles.
Q: Your recent work discusses the concept of “inter-regime doctrinal collapse” in data governance. Could you explain what this phenomenon entails and its implications for AI regulation?
A: Broadly speaking, my article explores how AI is challenging existing legal frameworks. Not in a literal sense, more fundamentally, by revealing that the doctrines and structures governing AI are not operating in clear, consistent, or principled ways. This has significant political, economic, and rule-of-law implications.
To make this concrete, I focus on data, especially data acquisition, since AI systems today require vast amounts of data to function effectively. How we regulate data directly impacts AI regulation. Because multiple legal fields like copyright law and privacy law apply to data, the regulatory landscape becomes complex. These fields have different rules and underlying goals. When the boundaries between copyright and privacy law blur, and their rules and rationales no longer remain distinct, the legal regimes can start to lose their structural integrity and effectively collapse into one another. I refer to this phenomenon as inter-regime doctrinal collapse.
A key point about the term: the word “collapse” might sound alarming, like a bridge falling down. But in this context, I use it descriptively. Whether this collapse is ultimately beneficial or harmful depends on what it enables, who gains power from it, and the broader political and legal consequences. So, it’s a phenomenon worth understanding. Moreover, it matters for AI regulation because data is a key input to develop and deploy AI systems, and we’re seeing doctrinal collapse with two regimes that govern data—information privacy law and copyright law.
Because I’m a big believer in showing, not just telling, I want to offer a concrete illustration that connects this point to AI regulation. Suppose a company scrapes data from the internet to train an AI model. Initially, the company claims that the data it uses is public and therefore not subject to privacy restrictions, because users voluntarily shared it. Simultaneously, it asserts that it’s not liable for copyright infringement because the data was publicly available. The term “publicly available” isn’t a legal term of art, but it appears frequently in AI disputes, litigation briefs, and public rhetoric. At the same time, the company refuses to disclose its training data, citing confidentiality and proprietary interests. Subsequently, the same company argues during litigation that user privacy requires non-disclosure despite previously denying privacy concerns because the data was shared voluntarily with a third party.
This example highlights how legal boundaries become fuzzy, even though copyright law and privacy law have very different doctrines and normative goals. In practice, what’s considered public versus proprietary, in copyright law, and public versus personal, in privacy law, become blurred as companies toggle between them. It becomes extremely difficult to determine which legal doctrine applies at any given moment. This is what I refer to as doctrinal collapse on the ground: the boundaries between privacy law and copyright law become indistinct, and the legal system’s structure begins to weaken. That connects back to AI regulation because the choices we are making about privacy law and copyright law regulate data—and because data acquisition is required for AI development, these choices about data regulation will affect AI governance.
Q : As you’ve just explained, your research suggests that the legal regimes governing data, privacy law, and copyright law are becoming increasingly blurred. What challenges does this pose for effective regulation of AI systems, and what institutional responses do you propose?
A: Here, I want to sharpen the political economy and rule of law stakes, and then consider some potential institutional responses.
First, collapse enables companies to manipulate the legal and social meanings of “public.” Not all companies are equally well-positioned to exploit the domains. Collapse tends to favor the “haves” – the well-resourced incumbents – by allowing them to acquire data at the expense of ordinary people or less well-resourced actors. In some cases, well-resourced incumbents have the money and influence to execute and leverage licensing agreements. In other cases, dominant platform firms are best-positioned to rely on broad user consent through privacy policies and terms of service.
Second, from a rule of law perspective, this fluidity allows private actors to switch between conflicting claims depending on what serves their interests, undermining legal predictability, coherence, and legitimacy. When laws become ambiguous and actors can switch between different legal regimes, it erodes public trust and the legitimacy of the legal system itself. This toggling creates a situation where the law no longer functions as a clear, predictable framework for accountability and justice, especially in the rapidly evolving context of AI.
Now, what might we do about this? I don’t believe that collapse itself is a problem we can solve. We can’t, and shouldn’t, try to create perfect clarity in the law or impose artificial boundaries between different fields of law. Instead, we should recognize that collapse becomes a problem when it undermines the law’s ability to govern effectively.
In the paper, I suggest some institutional responses. Some are incremental, focusing on adapting the current legal framework. In particular, we might draw on conflict of laws and empower courts as managers of collapse. For example, if a court is resolving a dispute where parties raise both privacy and copyright claims, the judge might insist on a rebuttable anti-switching presumption, saying, “You can’t assert mutually incompatible claims at different points in the lawsuit unless you provide a compelling reason to justify the switch.” These are strategies to manage the collapse without overhauling the entire legal system.
Other responses are more reformist, aiming to change the legal structure itself and make it harder to manipulate the lines between domains. Notably, we might adopt stronger privacy laws that make the initial relationship between copyright and privacy law less asymmetrical. I believe that reducing or eliminating the underlying weaknesses that lower the legal and social costs of privacy violations compared to copyright would decrease incentives for companies to exploit privacy loopholes and avoid copyright obligations.
Q: Let’s turn next to some of your other projects. In your paper on AI standards, you argue that standards are not neutral but have embedded politics. How can policymakers and regulators ensure that AI standards promote fairness and accountability rather than reinforce existing power structures?
A: Standards are tricky regulatory devices. One problem with standards, as I discuss in the paper, is that they tend to work much better in purely technical settings, like whether an outlet must have two or three prongs. But when we start dealing with socio-technical contested issues, such as fairness, accountability, or transparency, standard setting becomes much more difficult. I begin there because I think it’s important to be honest about that from the start. We can talk about how to improve standard setting, but standards are always political artifacts, in Langdon Winner’s sense of politics. They will inevitably reflect the power structures that create them.
For AI standards, one key step is to carefully consider whether a particular issue should be addressed through standard setting at all, or if it should be handled via public legislation or more binding legal procedures. If we decide that standard setting is appropriate, then we need to think about the relative influence of public versus private actors.
Another important aspect is ensuring ongoing deliberation and rethinking over time. I believe standards can sometimes be better than hard law because they can adapt more quickly and be nimbler. Building space for contestation and re-contestation is crucial. Without that, I worry that a powerful private actor could entrench a standard through market dominance, locking it in without democratic oversight or opportunities for re-evaluation. These chances to rethink things are vital, especially if what fairness means turns out to harm certain populations more than others, or if the initial requirements for transparency aren’t sufficient for outside parties to contest decisions made with the AI system, or if other problems with the standard emerge.
Q: Your work often emphasizes the importance of understanding the political and social context of technological standards. How can stakeholders ensure that AI governance frameworks are inclusive and reflect diverse societal values?
A: If I had a one-shot answer, I’d be selling it to the highest bidder, there’s no silver bullet. But I do think recognizing that technology is not a shiny, isolated object is a crucial first step. These tools are shaping our democracy and social future, and we must see that.
Technology is not neutral. Design choices, what data to use, how to define goals, how to align AI systems, or even whether to use AI at all, are deliberate decisions with real outcomes. The structure of our legal system also reflects choices; for example, how we regulate privacy or how lightly we regulate tech companies are policy decisions that shape society. My hope, and what I aim to help others see, is that these are choices and recognizing that empowers us. It means we’re not stuck; even without a perfect, one-size-fits-all solution, understanding these choices allows us to advocate for more inclusive and equitable governance.
Q: Your research also speaks to other aspects of AI regulation, such as the role of judicial decisions in shaping AI governance. How do you see the courts influencing the development of AI law, and what are the risks and benefits of relying on litigation as a form of regulation?
A: I see courts affecting AI law in many ways. It’s happening quite a lot. Sometimes, it’s direct, like the copyright lawsuits I discuss in AI and Doctrinal Collapse. Other times, it’s less direct, such as a First Amendment case that influences what policymakers believe is possible for AI regulation. Additionally, a company’s decision to settle a case rather than litigate can shape the regulatory landscape, depending on the outcome.
Whether this is good or bad is the key question. That’s why my piece is titled, “Do Cases Generate Bad Law?,” with a question mark. It’s a real question. Cases involve adversarial parties and concrete issues, where a harm that has already occurred. This can be beneficial because it helps focus attention on specific legal issues and concerns. Judges are often well-positioned to uncover detailed facts and understand how AI companies operate, which can be valuable.
However, whether courts make good or bad AI law depends a great deal on their interaction with legislators and regulators. AI cases are more likely to produce beneficial outcomes if they prompt legislative action to fill regulatory gaps and provide remedies for harms.
For example, there’s a recent lawsuit alleging a violation of Illinois’ Biometric Information Protection Act (BIPA). To establish a violation, there must be collection of biometric data, like faceprints or voiceprints, without prior informed consent. In this case, the complaint alleges that an AI company collected voiceprints to develop its system. If the case proceeds to discovery, it could reveal how these systems work, informing other plaintiffs, the public, and policymakers. It might lead Illinois legislators to update the law, if it turns out that they wanted to cover the data at stake and it is not covered. Or, if the voiceprints are covered by BIPA, the disclosures about how biometric information is used to create AI systems might inspire similar legislation elsewhere.
That said, there are risks to case-made AI law. Relying on litigation can mean that less tangible or emergent harms go unrecognized, either because there’s no legal cause of action or because the harm isn’t understood as such. It can also lead to spurious or costly litigation, which is especially problematic for startups.
Another concern is the concentration of cases in a few jurisdictions, which can result in a limited number of courts deciding complex social issues with nationwide implications. This lack of diversity in outcomes and limited access for plaintiffs is troubling. Therefore, I see litigation as an access-to-justice issue for individuals who may be left without a remedy for the negative impacts of AI systems. That is also why we need both litigation and legislation. It’s essential for functioning legislatures at both the state and federal levels to recognize how AI systems are affecting people, to realize the issues that cases are not likely to address, and to take action on these vital issues.
Q: Your research covers both AI law and information privacy law. In your view, how can the concept of the “Overton Window” be applied to improve the enforcement of privacy protections in the rapidly evolving digital landscape?
A: I’m not sure the Overton Window directly improves privacy enforcement. Instead, it helps reveal the range of actions that regulators believe are feasible at any given moment. What agencies do depend on internal norms, political will, resources, and external pressures from courts, industry, and social movements.
Applied to privacy, this means that privacy‑minded regulators can and should use broad legal authorities, like unfair and deceptive trade practices laws, to address emerging privacy and AI harms. They’re most likely to act where social consensus is already strong, such as with location tracking or children’s privacy. But the scope of enforcement ultimately turns on politics. Some FTC administrations have embraced a more expansive view of “unfairness,” enabling more aggressive interventions; others have taken a narrower approach.
Two additional points matter. First, enforcement requires resources. Privacy investigations are complex and time‑intensive. Without adequate funding and staffing, even strong legal tools won’t translate into meaningful action. Second, institutional design is crucial. When lawmakers create new privacy or AI rules, they need to consider whether agencies are insulated from political pressure and whether they have the capacity to act. Otherwise, even well‑intentioned laws risk being symbolic rather than effective.
Q: The broader question of how to regulate emerging technologies runs through your work. Given your expertise in algorithmic accountability and data governance, what are some practical steps that governments or organizations can take to enhance transparency and fairness in AI systems?
A: Much of my work is about reframing the problem rather than offering a single, crisp solution. It’s crucial to identify the assumptions underlying policymakers’ proposed paths forward. For example, governments or organizations should ask: What am I assuming about the law and the values I want to uphold? What about human actions, both by developers and end users? What assumptions are being made about the technology itself? By posing these threshold questions, we can reveal our underlying assumptions and develop interventions that better address complex human and technical interactions in specific contexts.
In addition, we should not latch onto just transparency, or just fairness, as the key to AI regulation. Although transparency and fairness are vital, focusing solely on those concepts can be limiting. Ultimately, the core issue is power, who controls the means to produce, refine, and deploy AI systems, and who has the voice to influence their governance. Overemphasizing just one aspect risks missing the relational dynamics at play. Recognizing these power relations is essential for meaningful accountability and equitable governance.
Q: As a member of the EPIC Advisory Board, a faculty affiliate at Harvard’s Berkman Klein Center, and an affiliated fellow at Yale Law School’s Information Society Project, how do collaborations across academia, civil society, and government influence the development of effective AI regulation?
A: The intersection of public and private sectors is vital for AI governance and my scholarship. First, collaborative governance, meaning active engagement between public and private actors is valuable in theory, but in practice, it carries risks. Without a strong, well-funded, and influential state, private actors can overshadow public voices, effectively replacing regulation with private interests. I believe that public regulation and democratic accountability are essential. They help ensure that AI development aligns with societal values, and contrary to some fears, regulation can foster innovation by guiding technology in lawful and ethical directions.
Second, the traditional public-private divide is increasingly strained by how social and governmental uses of AI are evolving. For example, in a forthcoming essay, called Clickwrap Accountability, I discuss how government agencies are using generative AI chatbots to provide non-binding advice on matters such as the Supplemental Nutrition Assistance Program (SNAP). This guidance often replaces in-person visits, government pamphlets, or static FAQ pages about benefits programs. These chatbot interactions are often mediated through private tools and rely on third-party terms of service, with limited accountability. When a chatbot provides advice, and there’s no formal government decision or official process involved, it doesn’t fall neatly within existing procedural due process frameworks. The current legal doctrine offers limited redress. At best, there are contract disclaimers or clickwrap style “agree” buttons for the end user. And because the state has sovereign immunity, the forms of contract or tort law redress that might be available in private law are not generally available in this public law context.
This example highlights a broader trend: the blurring of lines between public and private as AI tools become embedded in public services. Developments like this challenge our existing legal and regulatory frameworks and underscore the need for scholarship and policymaking to adapt to these new realities. We must rethink how accountability, transparency, and oversight are structured when government functions are mediated through private AI systems, and I’m working on these questions in several future projects.
Q: Looking ahead, what do you see as the most pressing legal or regulatory challenges in AI governance, and how should scholars and policymakers prepare to address them?
A: I see two major sets of challenges in AI governance. First, the friction between different legal regimes and values. AI systems sit at the crossroads of copyright, privacy, discrimination, consumer protection, and more. These areas don’t always align, and the trade-offs like balancing privacy with goals such as reducing bias or increasing transparency rarely have clean answers. Scholars can help by stepping outside disciplinary silos and examining how their preferred doctrines interact with others. Policymakers, meanwhile, should resist the urge for simple narratives. Effective regulation starts with mapping which bodies of law are implicated, where the gaps are, and what values are being traded off. There’s rarely a perfect solution, but there can be a principled one.
Second, we’re dealing with both known unknowns and unknown unknowns. Technological shifts like changes in data needs or the limits of so-called “scaling laws” could reshape incentives and alter how existing laws function. And then there’s the Collingridge dilemma: intervene too early, and policymakers won’t have the information they need, or too late, and harmful practices will already be entrenched. I tend to favor precaution where human interests are at stake, but with humility, flexibility, and mechanisms for revision as evidence evolves.
A final challenge is institutional capacity. Technology doesn’t inherently outpace law, yet knowing when to adapt and ensuring that agencies have the expertise to understand it is extraordinarily difficult. Robust governance will depend as much on institutional design as on the substance of any particular rule.
Q: What are some of the most exciting upcoming projects or research initiatives you are currently involved in or looking forward to?
A: I have several writing projects and initiatives that I’m looking forward to. First, I’m very excited about my forthcoming Clickwrap Accountability piece, as well as a forthcoming essay called The Supply Chain as a Circle: AI, Privacy, and People. These projects both focus on the users of technology and what existing law says or doesn’t say about these interactions. In the Supply Chain piece, I argue that the AI supply chain often overlooks the interactions between users and generative AI systems. If we don’t account for this interaction, then harm and responsibility tend to fall on end users, who often lack the knowledge or ability to prevent bad outcomes. They should be part of the regulatory calculus. In addition, I am developing a few other pieces, including one that examines relationships between administrative agencies and platform companies. All of these projects reflect my conviction that technological developments expose weaknesses in legal and regulatory frameworks and offer opportunities to re-examine institutions and doctrines.
Beyond scholarship, I’m engaging in conversations with EPIC about regulating AI chatbots and companion AI. I’m also involved in a Uniform Law Commission project on mental privacy, neural data, and cognitive biometrics, exploring potential model state legislation. This is a vital area that links privacy and AI. Think of the potential health benefits, like restoring hearing—but also the privacy risks of accessing our brains and most personal data.
Finally, I’m developing a new seminar called “Frontiers and Flashpoints in Tech Law,” which will cover cutting-edge issues like agentic AI, neural tech, and robotics. It aims to help students understand both the technology and the legal challenges. I am really excited for this course, and for all that undoubtedly lies ahead in tech law in the year to come.
Q: What do you think SSRN contributes to the world of modern research and scholarship?
A: I appreciate SSRN as a centralized place to find recent research. With so much information available online, it can be difficult to know where to look, so it’s incredibly helpful to have an open‑source community that makes research accessible and provides a clear, reliable destination for new work.
In addition, as a junior scholar, I’m especially grateful for SSRN as a research platform and as a way for my work to reach other scholars. I see it as a privilege to participate in these scholarly conversations, and I’m thankful for the infrastructure that makes it possible to sustain and expand them.
MORE ABOUT ALICIA SOLOW‑NIEDERMAN
Alicia’s scholarship explores how digital technologies disrupt traditional legal categories and institutional assumptions. Her influential work on inter‑regime doctrinal collapse shows how AI blurs the boundaries between privacy, copyright, and other legal domains, creating both regulatory challenges and opportunities for reform. She has written extensively on the politics of AI standards, the role of courts in shaping AI governance, the ways that inferences challenge privacy law on the books, and the need for institutional designs that can withstand uncertainty and technological evolution.
Her research has appeared or is forthcoming in leading journals, including the Stanford Law Review, Northwestern University Law Review, Harvard Journal of Law & Technology, Journal of Law & Innovation, and Southern California Law Review. A graduate of Harvard Law School, where she served as Forum Editor of the Harvard Law Review, Alicia has held fellowships at UCLA Law’s PULSE program and Harvard Law School, clerked on the U.S. District Court for the District of Columbia, and worked at the Berkman Klein Center for Internet & Society. She also serves on the EPIC Advisory Board and is a faculty affiliate at the Berkman Klein Center as well as an affiliated fellow at the Yale Law School Information Society Project, where she contributes to cutting‑edge conversations on AI regulation, mental privacy, and the future of public‑private governance.
You can see more work by Alicia Solow-Niederman on her SSRN Author page here.