DeepMind's CEO wants a referee for frontier AI

Google DeepMind CEO @demishassabis has called for the United States to establish a formal body to vet the world's most powerful AI models before they reach users, warning that artificial general intelligence is probably only a few short years away and that the window to build guardrails is closing fast.

A FINRA Model for Frontier AI

Hassabis is proposing an AI standards body modeled on FINRA, the private, industry-funded watchdog that polices Wall Street under SEC oversight. Frontier labs would initially share their models with the body voluntarily, up to 30 days before release, for safety testing that probes dangerous cyber, biological, and "deception" capabilities. Once the testing regime proves "effective and robust," formalization "could quickly follow," meaning frontier models would be required to pass before they could be deployed in the US market.

Hassabis envisions a majority-independent board stacked with Turing Award winners and other credentialed experts, alongside industry, government, and open-source representatives. The proposed body would need "substantial" funding to attract world-class technical talent and provide the necessary compute resources for large-scale testing, with that funding likely coming from industry. Hassabis says he wants the body operational before year-end.

The most striking element of the proposal is a brake mechanism. Hassabis says the body could be "ratcheted up" as risks grow, including "coordinating a slowdown in development among the Frontier Labs if deemed necessary." He says he has already briefed the Trump administration, rival labs, and European officials for months.

A Skeptical White House and an Awkward Funding Structure

The proposal lands at a complicated political moment. "There will not be an FDA for AI," White House AI adviser Sriram Krishnan told the Financial Times on July 3. He said the government favors a light-touch approach, warning that heavy regulation would slow innovation and weaken US competitiveness. The Hassabis framework, which deliberately avoids the language of government licensing, appears designed to thread that needle, placing oversight in industry hands rather than a federal agency.

But the self-regulatory model carries its own tensions. A body funded by the very labs it is meant to police faces an obvious conflict of interest, and the recent handling of model releases has done little to inspire confidence in ad hoc arrangements. Weeks before Krishnan's comments, the federal government made an "unprecedented intervention" forcing Anthropic to pull the latest version of its Mythos model, while also pausing the launch of OpenAI's 5.6. Hassabis cited those episodes as evidence that Washington needs a formal governance structure rather than improvised responses.

AGI capable of matching the human brain across a broad range of tasks is "probably only a few short years away," leaving society a "precious window" to establish oversight, according to Hassabis. Whether Washington can agree on a framework before that window closes remains the central open question.

Sources:
Axios: Google's Hassabis calls for new US-led global AI watchdog
CNBC: Google DeepMind chief calls for US to spearhead AI standards body
PYMNTS: White House adviser says Trump won't create FDA for AI