Google DeepMind chief Demis Hassabis on Tuesday called for the United States to spearhead a new standards body that would test advanced AI models for national security risks, including cybersecurity and biological threats. The proposal comes as the U.S.-China race to develop and deploy AI models intensifies and lawmakers weigh how to limit the adoption of Chinese systems by American companies.
Urgent action on AGI risks
Hassabis, a Nobel laureate, wrote in a post on X that "urgent action" was needed to address risks associated with artificial general intelligence (AGI) - the point at which AI matches or surpasses human intelligence. He said, "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."
A FINRA-style model for AI testing
Hassabis proposed a U.S.-led public-private partnership overseen by the federal government, modeled on the Financial Industry Regulatory Authority (FINRA). The body would require "substantial" funding - likely from industry - to attract top technical talent and provide computing resources for large-scale testing. Frontier AI labs would initially share models voluntarily for review up to 30 days before release, with the process becoming mandatory for U.S. market deployment once shown to be effective.
The testing would include specific checks for agentic AI systems, such as attempts to bypass safety guardrails or signs of deception. It would also ensure best practices like digitally watermarking AI-generated images and generating human-readable output tokens to understand model reasoning.
Growing calls for oversight amid industry tensions
The call for a standards body follows earlier efforts by tech leaders to shape AI governance. CNBC previously reported that Hassabis and Anthropic CEO Dario Amodei pushed for a U.S.-led AI coalition at a G7 meeting with world leaders, including President Donald Trump. OpenAI CEO Sam Altman also argued for a similar body in a Financial Times article earlier this month. The proposal comes as federal agencies and lawmakers increasingly focus on how to govern rapidly advancing AI systems, a subject at the heart of many AI for Government programs.
Regulation of leading AI models has become a flashpoint between the public and private sectors. In recent weeks, Anthropic negotiated with officials after the Trump administration temporarily imposed export controls on an advanced model. OpenAI also faced restrictions when the U.S. government initially requested limits on a new model's rollout.
China's competitive models raise the stakes
Recent model releases from Chinese companies, including DeepSeek and Z.ai, are seen as highly competitive with leading U.S. systems from Anthropic and OpenAI. They are gaining traction among American firms as AI costs rise. U.S. lawmakers are now considering measures to curb the growing adoption of Chinese AI models, which the State Department told CNBC raises "serious concerns."
Why this matters for Government
If a FINRA-style AI standards body moves forward, federal agencies and lawmakers will need to decide whether to support, fund, or develop alternative oversight mechanisms. The shift from voluntary to mandatory testing could set a precedent for how the U.S. government manages dual-use technologies, requiring new technical competencies within the public sector. For policy makers tasked with shaping these rules, understanding the technical nuances is critical, and resources like AI for Policy Makers can help build that expertise. The proposal also signals that industry leaders see federal intervention as inevitable, putting pressure on government institutions to act quickly and thoughtfully.
Your membership also unlocks:
