DeepMind CEO calls for FINRA-style AI regulator ahead of AGI
Google DeepMind's chief executive predicts artificial general intelligence is mere years away and urges the creation of a FINRA-style U.S. oversight body to manage escalating systemic risks.
Demis Hassabis has renewed his prediction that artificial general intelligence will arrive before the end of the decade, outlining a concrete regulatory framework to match the looming timeline. In a Tuesday blog post, the Google DeepMind CEO proposed establishing a U.S. Frontier AI Standards Body to evaluate advanced models prior to their public release.
The proposed entity would be modeled directly on the Financial Industry Regulatory Authority. Hassabis envisions a federally supervised, public-private partnership funded primarily by the AI industry and staffed by independent technical experts and open-source representatives. Pre-release testing under this framework could eventually become mandatory for the most capable systems.
Hassabis justified the regulatory push by warning that AI capabilities are advancing faster than society's ability to manage the risks. He pointed to existing cybersecurity threats and cautioned that future, more autonomous systems could introduce biological and nuclear dangers. “On the horizon, we will need robust safeguards to maintain control of increasingly agentic, recursively self-improving systems,” he wrote.
For investors and tech executives, the convergence of these accelerated timelines with specific regulatory blueprints signals a looming shift in how AI development is governed and funded. A FINRA-style oversight model essentially creates a new compliance layer, shifting the cost of safety auditing directly onto the industry's balance sheet. “The rapid progress we’re seeing in AI requires a new approach to testing frontier AI model capabilities that is dynamic, adaptable, and rigorous,” Hassabis wrote.
The proposal arrives as the industry's top executives align on a much shorter horizon for human-level AI. In January 2026, Anthropic CEO Dario Amodei predicted human-level AI could emerge within one to five years, warning governments were underestimating the pace. Last month, Amodei called for FAA-style safety rules, shortly after President Donald Trump signed an executive order establishing a voluntary review framework for advanced models.
Hassabis argued the U.S. has a narrow window to lock down these common standards before AGI materializes. “The future is not yet written, we must use this precious window before AGI arrives to shape this technology for the benefit of all humanity,” he wrote.
