DeepMind Co-Founder Proposes FINRA-Style AI Safety Regulator

A Wall Street model for AI oversight

Demis Hassabis, co-founder of Google DeepMind, has outlined a plan for a new U.S.-led AI safety regulator that could begin operations before the end of 2026. The proposed watchdog would test the world's most advanced AI models before release and eventually have authority to block systems deemed unsafe from entering the American market.

In a manifesto published Tuesday on Substack, Hassabis proposed structuring the body after the Financial Industry Regulatory Authority, the self-regulatory organization that oversees Wall Street broker-dealers. Like FINRA, the AI regulator would be federally overseen but largely funded by the industry it governs.

Hassabis leads Google's flagship AI lab after the tech giant acquired his London-based startup DeepMind in 2014. The proposed system would therefore regulate the same competitive landscape where Google faces off against OpenAI, Anthropic, and other frontier AI developers.

How the testing regime would work

Under the proposal, AI developers would initially submit models voluntarily up to 30 days before public release. The regulator would evaluate whether systems could enable cyberattacks, biological threats, or deceptive behavior.

These voluntary reviews could later become mandatory for models exceeding an evolving "frontier" threshold. The requirements would apply to both domestic developers and foreign or open-source systems seeking to enter the U.S. market.

Owen Larter, Google's senior director of frontier AI policy and global affairs, told Inc. that "the standards body can begin as an industry-led, voluntary organization, that in future gets formalized by statute or executive order."

Larter emphasized that AI companies should contribute technical expertise while keeping their influence "proportionate and balanced" against other board members. He clarified the rules would target only developers of frontier models, not the broader ecosystem of startups building applications on top of existing systems.

Why it matters

The proposal arrives as governments worldwide struggle to regulate AI development that outpaces traditional policymaking. A FINRA-style model could provide clearer, more consistent standards than ad hoc government interventions. But it also raises questions about whether leading AI companies should help write the rules governing their own competition—particularly when the proposal comes from an executive at one of the field's dominant players. The balance between industry expertise and regulatory independence will likely prove central to any debate over implementation.

Industry self-regulation or regulatory capture?

The FINRA comparison cuts both ways. Wall Street's self-regulator has provided decades of market oversight, but critics have long questioned whether industry-funded bodies can adequately police their own members.

For AI, the stakes extend beyond financial markets to questions of national security, public safety, and technological sovereignty. Whether a hybrid model can balance technical expertise with genuine independence remains an open question.

These details were first reported by Inc.