Google DeepMind CEO Demis Hassabis has proposed creating an independent standards body for frontier AI models, structured after the Financial Industry Regulatory Authority (FINRA). The idea is straightforward: build a self-regulatory organization funded by the AI industry itself, operating under federal oversight, that would voluntarily test the most powerful AI systems before they hit the market.
The initial proposal includes a 30-day review period before deployment of frontier models, giving the body time to evaluate risks before these systems go live.
What the FINRA model actually means for AI
FINRA is the private corporation that regulates broker-dealers in the US. It’s not a government agency, but it operates with SEC oversight and enforcement teeth. Brokerages fund it, follow its rules, and submit to its audits.
Hassabis wants to replicate that dynamic for AI. The proposed body would focus specifically on cutting-edge systems, the kind being developed by DeepMind, OpenAI, and Anthropic.
