Verify, Then Trust: What Lawyers Who Build Can Bring to the AI Rulebook

Guest Perspective | By Emily Cabrera

Florida’s AI fights are only part of this election. Nationally, three questions are in play: who gets to regulate AI, who is paying to shape the answer, and what voters are allowed to believe.

On preemption, the White House has pushed to override state AI laws, but an executive order cannot preempt state law on its own. Congress hasn’t acted, so state rules stay enforceable unless a court blocks them. Meanwhile, AI-aligned super PACs had spent about $55.7 million by late September, according to a New York Times analysis. And 31 states now regulate election deepfakes, including Florida, whose §106.145 requires a disclaimer on AI-generated political content made to injure a candidate or deceive voters about a ballot issue. Each of these fights comes down to technical lines. What counts as AI-generated? What must be disclosed? What has to be tested before anyone relies on it?

I saw why those lines matter this fall while mentoring Justin Tung of Jackson Walker in the LegalQuants AI Residency. Justin built a computer-vision model that pairs the images in past IP cases with the court’s analysis. According to LegalQuants, it took first place in Cohort 3. What made it credible was not the score but the discipline behind it. We narrowed the scope, tested it against simple baselines, and kept reminding ourselves that a model score is not a legal outcome.

That habit is verify, then trust, and policymakers need it too. Lawyers who have built and tested AI tools can tell where a rule’s technical language will hold up and where it won’t. You don’t need to be an engineer to build. Start small, with a real problem you understand. LegalQuants Cohort 4 begins November 1, and I’ll be mentoring again.