Trump Administration Weighs a FINRA-Style Watchdog to Vet Frontier AI Models
A Treasury-backed proposal would create an industry-funded regulator reporting to the SEC, running mandatory 30-day safety reviews of frontier models before public release.
Treasury Secretary Scott Bessent has helped develop a proposal for an independent AI regulator modeled directly on FINRA, the private, industry-funded body that polices Wall Street under SEC oversight, according to a Bloomberg report from July 17. The new agency would run mandatory 30-day safety evaluations on frontier models — screening for cyber, biological, and deceptive-behavior risks — before labs can ship them publicly.
Like FINRA itself, major AI labs would fund the agency directly rather than taxpayers footing the bill, while the SEC retains ultimate oversight. That structure — industry money, government backstop — is designed to sidestep the two failure modes Washington keeps hitting: an agency too slow and underfunded to keep pace with model releases, or one so politically exposed that every decision becomes a fight. The plan is currently under review by White House Chief of Staff Susie Wiles, and Bloomberg is explicit that this is a proposal, not policy — voluntary, unfinished, and open to being rewritten or shelved entirely.
The proposal didn’t emerge in a vacuum. It follows Silicon Valley complaints about the ad-hoc way the administration has already been gating model releases this year — OpenAI was reportedly asked to gate its GPT-5.6 release, and Anthropic spent weeks negotiating after Claude Mythos 5 and Fable 5 access was blocked over unspecified national security concerns before being reinstated. A standing, rules-based review body would replace those one-off interventions with a predictable process labs can actually plan a release calendar around.
There’s an unusual convergence here worth noting: Google DeepMind CEO Demis Hassabis published a manifesto on July 14 independently proposing almost the identical structure — a U.S.-led, FINRA-style standards body for frontier AI. When a Treasury-linked proposal and a leading lab’s CEO land on the same regulatory model within days of each other, that’s a stronger signal the idea has real momentum than either alone would suggest.
If adopted, this would be the first time frontier AI releases face a standing regulatory gate rather than the executive order’s current voluntary early-access framework. For labs, that trades unpredictability for a fixed 30-day compliance cost on every major release — a tradeoff most seem increasingly willing to accept over the status quo.