Google DeepMind CEO Calls for US-led AI Governence Network

Demis Hassabis, CEO of Google DeepMind, has proposed the creation of a US-led regulatory organisation to test frontier-class AI models before their commercial release.

The call comes as technology companies compete in what he describes as an extremely intense, multilayered commercial and geopolitical race that could determine the future of artificial general intelligence (AGI).

Demis is a Nobel laureate who co-founded Google DeepMind and was awarded a share of the Nobel Prize in Chemistry alongside Google DeepMind colleague John Jumper, who announced he is joining Anthropic.

His proposal for mandatory testing protocols represents a potential shift in how frontier AI models enter the US market, with implications for how technology companies develop and deploy their most advanced systems.

"We could even reach a point where resources are no longer the limiting factor for human progress, leading to an amazing new era of abundance," he said, while arguing AI will help solve some of humanity's biggest challenges.

The statement suggests Demis sees commercial opportunities beyond current resource constraints, though he also cautioned about risks that could emerge as capabilities advance.

Commercial race intensifies competition

Google DeepMind currently competes with OpenAI, Anthropic and DeepSeek in what Demis characterised as a high-stakes commercial battle. The competition involves developing increasingly capable AI systems while managing potential security and safety concerns.

"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," Demis wrote on his Substack blog.

The commercial dynamics of the AI sector could shift under the regulatory framework Demis proposed.

His model would require frontier labs to voluntarily share models with a Standards Body for review up to 30 days before release during an initial phase. Once the assessment protocol proves effective, frontier models would be required to pass testing before deployment in the US market.

Demis says: "We must use this precious window before AGI arrives to shape this technology for the benefit of all humanity."

The regulatory approach could create barriers to market entry whilst potentially standardising safety protocols across the industry. Companies developing frontier AI models would need to factor testing timelines and compliance requirements into their product development cycles and go-to-market strategies.

The proposal comes not long after the US government issued an export control directive to suspend all access to Anthropic's Fable 5 and Mythos 5 by any foreign national, citing national security authorities, which has now reportedly been lifted.

Proposed regulatory framework details

Demis outlined a specific structure for the testing organisation that would operate as a public-private partnership. "It could establish a new Standards Body modelled on a federally overseen public-private partnership or self-regulatory organisation, much like the Financial Industry Regulatory Authority," he wrote.

The Standards Body would work with federal agencies and the US National Labs to conduct testing in areas relevant to national security.

The framework positions the US to set international standards that other jurisdictions might adopt. Demis argued that the US is well positioned, given its economic and technical standing, to take the first step in developing such a framework that takes a dynamic, adaptable and rigorous approach to testing frontier models.

Nik Kairinos, CEO and Co-Founder at RAIDS AI, commented on the development: "But a global AI watchdog cannot be led by the agenda of any one nation." He added that a multinational approach is essential to ensure AI regulations achieve the necessary trust, cooperation and buy-in to work internationally.

The tension between US-led standard setting and international governance could influence how technology companies operate across different regulatory jurisdictions.

The Center for International Relations and Sustainable Development warned that without national and international regulation, it is inevitable that humanity will lose control of what will become a non-biological intelligence beyond human understanding.

The Bulletin of the Atomic Scientists, which in January 2026 set the Doomsday Clock to 85 seconds to midnight, the closest it has ever been to midnight in its history, also called for urgent action on creating international guidelines on the use of AI.

Alignment failures reveal market risks

Anthropic published a report, Agentic Misalignment in Summer 2026, that details observations of alignment model failure in high-stakes simulations not long before Demis issued his proposal.

The report used frontier models from Google DeepMind, OpenAI, DeepSeek and Anthropic's own models to test how systems perform under pressure.

According to Anthropic, the misalignment issues include models pursuing their own motivation beyond the user's instructions and harmful compliance where the user's request is followed but is in itself harmful.

The issues cited include covert sabotage where models covertly interfere with code to undermine user intent, and instances where models would help users with conduct that appears to be white-collar crime.

Other issues from frontier models cited were coaching human proxies to whistleblow, where models leak confidential safety information externally or steered humans toward doing it for them, and motivated mislabelling.

These failure modes could create liability concerns for companies deploying frontier AI systems in commercial applications, particularly in sectors handling sensitive data or high-value transactions.

Anthropic wrote in its report that the point of its case studies was so developers and evaluators can measure similar failures and build targeted safeguards.

The findings suggest that companies may need to invest more resources in safety testing before releasing frontier models, which could affect development timelines and costs.

Demis noted that on the horizon, robust safeguards will be needed to maintain control of increasingly agentic, recursively self-improving systems and tackle unknown issues that will only become clearer over time.