Google DeepMind CEO Demis Hassabis called for the U.S. to lead a standards body overseeing new AI models and assessing national security risks including cybersecurity and biological threats, in an article posted on X on Tuesday. The Nobel laureate said urgent action was needed to address risks associated with artificial general intelligence (AGI), noting that frontier models already pose cybersecurity challenges and that nuclear and bio risks may soon emerge as AI capabilities advance. The call comes amid growing regulatory tensions, with Anthropic and OpenAI recently facing temporary export controls and rollout restrictions from the Trump administration.
Hassabis Proposes U.S.-Led Public-Private Partnership for AI Oversight
Hassabis proposed a U.S.-led public-private partnership overseen by the federal government as a solution to tackle AI threats. He said the U.S. was well positioned to lead in developing an AI framework given its economic and technical standing. The proposed body would be modeled on a federally overseen public-private partnership or self-regulatory organization, similar to the Financial Industry Regulatory Authority (FINRA), with a board including independent leading technical experts and open-source representatives.
The standards body would require substantial funding to attract world-class technical talent and provide necessary compute resources for large-scale testing, with funding likely coming from industry, according to Hassabis. Frontier labs would initially voluntarily share models with the body for review up to 30 days before release, before becoming mandatory for deployment in the U.S. market after being shown to be effective.
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 comments come a month on from Hassabis and Anthropic CEO Dario Amodei calling for a U.S.-led coalition to shape rules and standards around AI at a G7 meeting with tech leaders and heads of state that included President Donald Trump, according to sources who spoke to CNBC. OpenAI's Sam Altman also called for a similar body in an article published by the Financial Times earlier this month.
![chart visualization]()
Chinese AI Models Gain Traction as U.S. Lawmakers Consider Curbs
Calls for greater regulatory oversight come as the race between the U.S. and China to develop and deploy AI models heats up. Recent model releases from Chinese companies, including DeepSeek and Z.ai, are seen by many as highly competitive compared to leading frontier systems from Anthropic and OpenAI, and are gaining traction among U.S. companies as AI costs rise.
U.S. lawmakers are currently considering how to curb the growing adoption of Chinese AI models by homegrown companies, which the State Department told CNBC raises serious concerns. In recent weeks, Anthropic was locked in negotiations with officials after the Trump administration temporarily imposed export controls over an advanced model. OpenAI also faced restrictions as it was initially requested by the U.S. government to limit the rollout of a new model.
![Chinese AI models are gaining ground with U.S. companies]()
FAQ
What did Demis Hassabis propose on Tuesday?
Demis Hassabis, Google DeepMind CEO, called for the U.S. to lead a standards body overseeing new AI models and assessing national security risks including cybersecurity and biological threats in an article posted on X on Tuesday.
How would the proposed AI standards body operate?
The proposed body would be modeled on a federally overseen public-private partnership similar to FINRA, with a board including independent technical experts and open-source representatives. Frontier labs would initially voluntarily share models for review up to 30 days before release, before becoming mandatory for U.S. market deployment after being shown to be effective. Funding would likely come from industry to attract technical talent and provide compute resources for large-scale testing.
