'Urgent Action' Needed as AI Capabilities Accelerate
Google DeepMind CEO Demis Hassabis — a Nobel laureate and one of the most influential figures in artificial intelligence — has called for the creation of a U.S.-led standards body to test frontier AI models for national security threats before they reach the public.
In an article posted on X on Tuesday, July 14, Hassabis said "urgent action" was needed to address risks associated with artificial general intelligence (AGI). "We've already seen the challenges frontier models pose for cybersecurity, and other threats including nuclear and bio risks may soon emerge as capabilities continue to advance," he wrote.
Modeled on FINRA
The proposed "Frontier AI Standards Body" would be structured as a public-private partnership overseen by the federal government, modeled on the Financial Industry Regulatory Authority (FINRA) — the U.S. body that regulates brokerage firms and exchange markets.
Key elements of the proposal:
- Voluntary to Mandatory: Frontier labs would initially share models voluntarily up to 30 days before public release. After proving "effective," submission would become mandatory for deployment in the U.S. market. - National Security Testing: The body would assess models for capabilities in cybersecurity attacks, biological weapons development, and ability to bypass safety guardrails. - Industry Funding: The organization would need "substantial" funding to attract world-class technical talent and provide the necessary computing resources for large-scale testing — likely funded by the industry itself. - Open Source Representation: The board would include independent technical experts and open-source representatives.
"Specific agentic AI tests could look for attempts to bypass safety guardrails or signs of deception, and ensure best practices, such as digitally watermarking AI-generated images and generating human-readable output tokens to understand model reasoning," Hassabis said.
The Growing Consensus
Hassabis's proposal follows a broader push from AI industry leaders for formal oversight. Sources previously told CNBC that Hassabis, alongside Anthropic CEO Dario Amodei, called for a U.S.-led coalition to shape AI rules at a G7 meeting with tech leaders and heads of state including President Donald Trump. OpenAI's Sam Altman also made a similar call in a Financial Times op-ed earlier this month.
The growing consensus comes as:
- Export controls escalate: The Trump administration recently imposed temporary export controls on an advanced Anthropic model. - OpenAI faced restrictions: The U.S. government initially requested OpenAI limit the rollout of a new model. - Chinese models gain ground: Systems from DeepSeek and other Chinese firms are increasingly competitive with U.S. frontier models, raising national security concerns.
The U.S.-China Dimension
The proposal lands at a moment of intensifying U.S.-China AI competition. Chinese AI models are gaining adoption among U.S. companies as costs rise for leading American systems. The State Department has told CNBC this trend raises "serious concerns."
Hassabis argued the U.S. was uniquely positioned to lead such an effort "given its economic and technical standing." A formal standards body could give developers clearer requirements while allowing governments to compare systems using shared methods.
Challenges Ahead
However, such an organization would face difficult questions about: - Independence from the industry it regulates - Enforcement authority and international participation - Whether testing requirements could favor large incumbents over smaller labs - How to keep pace with rapidly advancing capabilities
The debate over AI regulation is shifting from whether oversight is needed to who will lead it — and how fast they can move.
