chief of Google The AI division called on the United States to spearhead a standards body that would oversee new AI models and assess national security risks, including cybersecurity and biological threats.
Google DeepMind boss and Nobel laureate Demis Hassabis said in an article published in X on Tuesday that “urgent action” is needed to address the risks associated with artificial general intelligence (AGI), or AI, to the point where it matches or exceeds human intelligence.
“We are already seeing the challenges that frontier models pose to cybersecurity, and as capabilities continue to advance, other threats, including nuclear and biorisks, may soon emerge,” he said.
Hassabis proposed a U.S.-led public-private partnership with federal oversight as a solution to help address these threats. The White House, State Department, and Commerce Department have been contacted for comment.
The comments came a month ago when Hassabis told CNBC that he and Anthropic CEO Dario Amodei called for a U.S.-led coalition to shape rules and standards around AI at a G7 meeting with technology leaders and heads of state, including President Donald Trump. OpenAI’s Sam Altman also called for a similar group in an article published in the Financial Times earlier this month.
AI standards organization
Despite growing calls for an AI watchdog among industry leaders, regulation of major AI models is increasingly a point of contention between the public and private sectors.
Anthropic has been locked in negotiations with officials in recent weeks after the Trump administration imposed temporary export restrictions on advanced models. OpenAI also initially faced restrictions due to requests from the U.S. government to limit the deployment of new models.
Hassabis said the United States is in a position to lead the development of AI frameworks “given its economic and technological position.”
“A new standards body could be created, much like the Financial Industry Regulatory Authority (FINRA), modeled on a federally supervised public-private partnership or self-regulatory organization, with a board that includes leading independent technology experts and open source representatives,” he added. FINRA regulates U.S. securities firms and exchange markets
Hassabis said the proposed organization would require “substantial” funding “to attract world-class technical talent and provide the computational resources needed for large-scale testing.” He added that funding from industry was “likely”.
Frontier Labs will initially be required to voluntarily share the model with organizations for review up to 30 days before launch, and then roll it out to the U.S. market once it proves to be “effective.”
“Specific agent AI testing can look for attempts to circumvent safety guardrails or signs of deception, and ensure best practices such as watermarking AI-generated images or generating human-readable output tokens to understand model inferences,” Hassabis said.
US vs. China
As competition between the U.S. and China intensifies over the development and deployment of AI models, calls for stronger regulatory oversight are growing.
Recently released models from Chinese companies like DeepSeek and Z.ai are seen by many as more competitive compared to major frontier systems like Anthropic and OpenAI, and are gaining traction among U.S. companies as AI costs rise.
As a result, U.S. lawmakers are now considering ways to curb the growing adoption of Chinese AI models by domestic companies, which the State Department told CNBC raises “grave concerns.”
