The Hook
Clearview AI built its product by scraping photographs from the public internet — social media profiles, news sites, anywhere a face appeared — and assembling a database it claimed contained over 10 billion images (the ACLU complaint cited more than three billion faceprints at its 2020 filing). It sold search access to law enforcement and, for a period, to private entities. Illinois had a law that said you needed consent before capturing a faceprint.
The Question
When the industry argument for federal AI preemption is that state laws create a "patchwork," what does the patchwork's track record look like?
The Paper Trail
Illinois' Biometric Information Privacy Act (2008) requires informed written consent before a company collects a biometric identifier, including a faceprint, and gives individuals a private right of action — ordinary people can sue, not just regulators. The ACLU and its partners sued Clearview AI on behalf of survivors of domestic violence and sexual assault, sex workers, and undocumented immigrants — people for whom public face-recognition databases create direct physical risk.
The case ended in a binding consent order, dated May 11, 2022: Clearview permanently barred from selling database access to private entities nationwide, and barred for five years from selling to Illinois state and local government including police. A single state law, enforced by private plaintiffs, changed the company's national behavior — not just its Illinois behavior.
The Synthesis
The preemption argument is that inconsistent state rules impose compliance costs and create uncertainty. That is true. The argument omits that those same state rules — written by elected legislatures, enforced by people directly harmed — produced the most concrete AI-rights outcome in the U.S. public record. The preemption campaign described in the Ballot Moat story is trying to override the mechanism that produced this result. That is worth noting before accepting that federal uniformity is a neutral goal.