---
title: "DeepMind's Demis Hassabis Calls for a US-Led Global AI Watchdog"
description: "Google DeepMind chief executive Demis Hassabis has proposed a new safety body, led by the United States and modeled on Wall Street's self-regulator, to vet the most powerful AI systems before release. He wants it running before year-end, framing AI oversight as an urgent economic and security question."
category: "Tech"
category_url: https://boursel.com/category/tech
author: "Sofia Marchetti"
published: 2026-07-14T13:28:00.000Z
updated: 2026-07-14T13:28:00.000Z
canonical: https://boursel.com/article/deepmind-s-demis-hassabis-calls-for-a-us-led-global-ai-watchdog
tags: ["artificial-intelligence", "ai-regulation", "deepmind", "governance"]
---
# DeepMind's Demis Hassabis Calls for a US-Led Global AI Watchdog

Google DeepMind chief executive Demis Hassabis has proposed a new safety body, led by the United States and modeled on Wall Street's self-regulator, to vet the most powerful AI systems before release. He wants it running before year-end, framing AI oversight as an urgent economic and security question.

Demis Hassabis, the Nobel-winning chief executive of Google DeepMind, is calling for a new watchdog to police the most advanced artificial-intelligence systems, arguing that the technology is now moving faster than the rules meant to govern it. In a framework published on July 14, [Hassabis proposed a US-led body that would vet frontier AI models before they reach the public and, he says, ideally be operational before the end of the year, according to Axios](https://www.axios.com/2026/07/14/demis-hassabis-ai-regulation-google-deepmind).

## A regulator modeled on Wall Street's

Hassabis is not asking governments to build and run the regulator themselves. Instead he suggests a standards body modeled on FINRA, the private Financial Industry Regulatory Authority that oversees US brokerages under the eye of the Securities and Exchange Commission. [As Axios reported, the proposed board would draw on leading researchers, including Turing Award winners, alongside industry, government and open-source representatives](https://www.axios.com/2026/07/14/demis-hassabis-ai-regulation-google-deepmind).

The mechanics would start out voluntary. AI labs would submit their most capable "frontier" models up to 30 days before a planned release for safety testing aimed at spotting dangerous capabilities, in areas such as cybersecurity, biology and the potential for the systems to deceive. The rules, in Hassabis's telling, would apply to all frontier-class models regardless of where they are built or whether they are open or closed source. If the regime proved effective, he suggested, it could later become mandatory, gating access to the US market.

## The IAEA and CERN comparisons

Hassabis reached for two well-known international institutions to make his case: the International Atomic Energy Agency, which monitors nuclear material, and CERN, the European physics laboratory, as a model for shared, collaborative research. The analogy is a common one in AI-policy debates, but it is not a clean fit. As the national-security outlet Lawfare has noted, [the IAEA lacks direct enforcement power and depends on great-power consensus, a weakness that geopolitics can exploit](https://www.lawfaremedia.org/article/do-we-want-an--iaea-for-ai). AI adds its own complication: there is no scientific consensus on when, or whether, the most extreme risks would ever materialize.

## Why this lands as a business story

For companies and investors, the framing matters as much as the specifics. A senior figure at one of the world's leading AI labs is arguing that oversight of frontier models should be organized, centralized and, crucially, anchored in the United States, a stance that carries competitive and geopolitical weight. How AI is governed, and by whom, will shape which firms can ship which products, in which markets, and on what timeline.

It is also, for now, a proposal rather than policy. Google, OpenAI and other labs have floated overlapping governance ideas that agree on risk-based testing of powerful models but diverge sharply on how much formal government authority should sit behind it. Hassabis's contribution pushes a specific structure into that debate and attaches a deadline to it. Whether regulators, rivals and other governments accept an American-led body as the referee is the open question.

## Sources

- [Google's Hassabis calls for new US-led global AI watchdog 'before year end'](https://www.axios.com/2026/07/14/demis-hassabis-ai-regulation-google-deepmind)
- [Do We Want an IAEA for AI?](https://www.lawfaremedia.org/article/do-we-want-an--iaea-for-ai)