Google DeepMind CEO Demis Hassabis is heading to Washington with a pitch that sounds surprisingly familiar to anyone who’s watched financial regulation evolve: let the industry police itself, but give the feds a seat at the table.
Hassabis announced on July 14 a proposal to create a US-led independent standards body for frontier AI, structured as a self-regulatory organization modeled after FINRA, the Financial Industry Regulatory Authority that oversees broker-dealers on Wall Street. The idea is straightforward. AI labs would voluntarily submit their most powerful models for independent review up to 30 days before release, with the possibility of mandatory assessments down the road.
The FINRA playbook, applied to AI
FINRA is funded by the financial industry it regulates, operates under SEC oversight, and has the authority to fine firms and ban individuals. It’s not a government agency, but it carries real enforcement teeth. Hassabis wants something analogous for AI: industry-funded, federally overseen, capable of conducting pre-release testing on frontier models.
The framework would apply to both open-source and closed-source AI models. Initially, participation would be voluntary. Labs could opt in, submit their models, get a review, and presumably use the stamp of approval as a marketing asset.
