“Clearview” into the Shortcomings of the Data Privacy Paradigm
The recent Clearview AI settlement showcases how our legal system can already support collectively-owned AI, highlighting the immense potential of getting the data ownership conversation right.
The facial recognition startup, accused of invasion of privacy in a class-action lawsuit, has agreed to a settlement with a twist: Rather than cash payments, it would give a 23% stake in the company to Americans whose faces are used in its AI. Clearview scraped billions of photos from the web and social media sites such as Facebook, LinkedIn, and Instagram to build a facial recognition app used by thousands of police departments, the Department of Homeland Security, and the FBI.
This case demonstrates that the legal system can support the concept of collectively owned AI. However, the settlement is still a sub-par outcome. While several regulators have demanded that Clearview remove peoples’ faces from its AI, the company admits it can’t feasibly do that. This ex-post approach to privacy is problematic, and it stems from a fundamental misstep by policymakers and regulators, who focus too much on a “right to privacy” and too little on the individual’s inherent property rights in their data.
Hohfeldian Rights Analysis helps illustrate this subtle but essential distinction.
When we say that people have a “right to privacy,” we’re actually saying that we have “immunity” from specific actions regarding our data. While “immunity” is a powerful concept, it doesn’t fully encapsulate the rights we can assert regarding our data. We can only say that companies can’t do certain things with our data without our consent.
When we start talking about data as property, the classification shifts. In a Hohfeldian sense, this transition moves our rights from immunity to power. This means we can change legal relationships as they relate to our data. Practically, this translates to having the power to financialize our data, control where it is used, and gain income from it, signifying a more comprehensive exercise of our rights than merely preventing unauthorized use.
The rhetoric that keeps data rights in the realm of immunity (i.e., privacy) limits the full extent of what can be achieved. Discussing data ownership broadens the scope of the “right,” allowing individuals to fully leverage the power of their data. This includes determining the technology that uses their data, using data as financial collateral, and earning a stake in the resulting innovations.
The Clearview case brings this issue to the forefront. A system grounded in data ownership would have resulted in a different analysis of the issues. If we assume proprietary interest in personal data, a settlement involving owning a piece of the end product wouldn’t be so novel. Principles of common law or equitable tracing could apply.
From a user’s perspective, a system that prioritizes ownership would also emphasize the need for a bona fide transaction to transfer property. This would mean that individuals would need to freely choose whether to contribute their data to the model in the first place, with full awareness and control over the potential benefits and risks.
In the specific case of Clearview, facial recognition technology might be valuable for safety and identity verification for some. However, participation in such technologies might be undesirable for others, particularly marginalized groups wary of surveillance and discrimination.
As a queer person of colour, I would be cautious about contributing my data to Clearview due to potential misuse in government and policing contexts. But others might wish to do so. The current legal framework, focused on “privacy rights”, does not deal with diversity of preference well. The standard approach is to offer confusing, broad-sweeping general disclosures that force me to waive my legal immunity from data mistreatment. Put simply, if data were treated as property, I would have more control and could negotiate how my data is used and what I receive in return.
A genuinely empowering system would recognize data as a tradable asset and personal property, enabling individuals to decide how their data is used and ensure they benefit from its value. This perspective is crucial as AI becomes more integrated into our lives. As a core ingredient of the AI stack, we share incredible decentralized power to shape emergent technology with our data. Therefore, the structure of data rights will be critical in determining whether we can prioritize innovation that adds value for humanity.
At Vana, we are pioneering a vision where individuals have the right to their data and can collectivize to create collectively owned datasets and AI. At its core, Vana is user-owned infrastructure for a data-powered economy, owned by its participants rather than centralized entities. What makes it powerful from a technical perspective is equally important from a societal one. This approach shifts the paradigm towards decentralized innovation and collective ownership.
The Clearview AI case illustrates how collective ownership can emerge but also reveals the need for better first principles in data rights. Recognizing data as a personal property right, rather than just an immunity, is fundamental.