Demis Proposes an AI Version of FINRA
chris_j_paxton · x · 2026-07-15
Demis Hassabis proposed a governance framework akin to FINRA for AI: US standards bodies would set baseline thresholds, and models reaching these thresholds would be classified as "Frontier-class".
The post outlines key points: labs submit models to the agency about 30 days before release; initially voluntary, this could evolve into mandatory pre-release reviews in the US; evaluation criteria are updated quarterly, and the agency might even build its own reserved test sets. The author cautions that while making benchmarks legally binding is a clear idea, it also faces challenges like regulatory capture, test gaming, and implementation complexity.
Related event: Hassabis Says AGI May Be Years Away, Proposes Frontier AI Governance(23 posts)→
More from AGI Musings
- The Evolution of LLM Business Models: Selling Outcomes Over Tokens — yacineMTB · 2026-07-22
- Bindu Reddy says GPT-6 is coming soon, with Alibaba, DeepSeek and Kimi close behind — bindureddy · 2026-07-22
- Bindu Reddy says the industry still lacks a way to train 20T models and scale post-training RL — bindureddy · 2026-07-22
- Advanced AI Models Are Becoming Impossible to Plug and Play — emollick · 2026-07-22
- AI suggested a better composition, and that made one user uneasy — Sydde · 2026-07-22
- The Thimble and the Waterfall: AI's Data Bottleneck and Feedback Loops — dyamins · 2026-07-22