Google DeepMind Chief Urges Washington to Lead Global AI Safety Standards

The head of Google's AI division has called on the United States to establish a federal body responsible for overseeing advanced artificial intelligence models and evaluating national security risks - especially in cybersecurity and biosecurity. Demis Hassabis, a Nobel laureate and CEO of Google DeepMind, warned in a post on X Tuesday that "urgent action" is needed as artificial general intelligence - where machines match or exceed human thinking - draws closer. "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.

Hassabis proposed a public-private partnership overseen by the federal government, modeled on the Financial Industry Regulatory Authority, which regulates U.S. brokerage firms. The proposed standards body would be funded largely by industry to attract top technical talent and provide the computing power for large-scale testing. Frontier labs would initially submit models voluntarily for review up to 30 days before release, with mandatory compliance later if the system proves effective. Specific tests would check for attempted bypass of safety guardrails, signs of deception, and ensure practices like digital watermarking for AI-generated images.

The call comes amid growing tension between tech leaders and regulators. In recent weeks, Anthropic faced export control restrictions from the Trump administration over an advanced model, and OpenAI was asked to limit a new model's rollout. Hassabis argued the U.S. is uniquely positioned to lead given its economic and technical strength. His proposal follows similar appeals from Anthropic CEO Dario Amodei and OpenAI's Sam Altman, as competition with China intensifies. Chinese models from companies like DeepSeek and Z.ai are gaining traction among U.S. firms, prompting lawmakers to consider curbs on adoption over national security concerns.

Source: CNBC