Demis Hassabis, co‑founder of DeepMind, has published a policy proposal for regulating artificial general intelligence (AGI). His core suggestion is that the US government should create a framework to test frontier AI systems for emergent capabilities via a Standards Body — a federally overseen public‑private partnership or self‑regulatory organisation modelled on bodies such as the Financial Industry Regulatory Authority (FINRA).
Responsibilities of the Standards Body
According to Hassabis, the Standards Body would develop assessment protocols and work with appropriate federal agencies and the US National Labs to conduct testing in areas relevant to national security. This testing infrastructure would also help define the criteria that make a model a "Frontier Model." Labs developing such models would be "encouraged" to adopt best practices, including publishing details about their systems, investing in cybersecurity, and vetting personnel.
Start voluntary, then consider legal formalisation
Hassabis proposes a phased approach: initially, Frontier Labs would voluntarily share models with the Standards Body for review up to 30 days before release. Once the assessment protocol is shown to be effective and robust, formalisation — meaning potential legal requirements — could follow rapidly.
Why this matters
The proposal is notable because it pulls together several de facto consensus positions that have emerged across the AI industry: powerful AI systems should be subject to third‑party testing, and those third parties should have some connection to a regulator such as the US government.
The idea also aligns with recent US policy trends. Both the Trump administration's executive order on AI and processes developed after the Anthropic export controls episode moved toward collaborative assessment methods between government and industry to evaluate AI capabilities and national security risks.
Google and comparable proposals
Hassabis's piece is also significant because Google has not often published such specific policy prescriptions; the proposal is unusually detailed from that company and sits alongside spiritually similar, though not identical, proposals from other organisations such as Anthropic.
Source: "A Framework for Frontier AI and the Dawning of a New Age" (Demis Hassabis, post on X).
