As Google's artificial intelligence (AI) institutional sector chief argued that, with artificial general intelligence (AGI) looming, a new regulator led by the United States is needed.
Demis Hassabis, Nobel chemistry laureate and Google DeepMind Chief Executive Officer (CEO), said on the 14th (local time) on the social network service X (formerly Twitter) that "there are only a few years left until AGI arrives." Known as the father of the Go AI "AlphaGo," which defeated 9-dan Lee Sedol in 2016, CEO Hassabis jointly won the 2024 Nobel Prize in chemistry for developing the AI program "AlphaFold," which predicts protein structures.
CEO Hassabis emphasized, "The scale of this technology's impact is unprecedented and will likely unfold at a speed 10 times greater than the Industrial Revolution," adding, "AGI is closer to the discovery of electricity or fire than to the internet or mobile."
He said the uncertainty and risks from the advent of AGI are significant and that we should maintain "cautious optimism," arguing for the creation of a relevant regulatory body. Hassabis, a British citizen, noted, "Given its economic and technological standing, the United States is well positioned to take the first step in establishing such a framework," expressing the view that the U.S. should lead it.
He pointed to the Financial Industry Regulatory Authority (FINRA), a private entity that regulates Wall Street under the oversight of the Securities and Exchange Commission (SEC), as a model. He said a new body should evaluate AI models and, in cooperation with federal agencies and others, be responsible for testing them in national security–related areas such as cybersecurity and biological risks.
To that end, he proposed that, at first, cutting-edge AI labs voluntarily submit their models to a standard body 30 days before release for review, and that if such evaluation procedures prove effective, submission should become mandatory. As advantages of this approach, he cited focusing on technological progress while simultaneously supporting innovation and encouraging responsible behavior.
He also stressed that if a U.S.-led body is established in this way, it would lead to setting international standards for AI. Hassabis' proposal is seen as related to the U.S. Department of Commerce recently issuing export-ban guidance that, for security reasons, temporarily barred foreign-user access to Anthropic's AI model "Claude Mythos," among others. OpenAI also went through a process of pre-releasing its latest AI model, "GPT-5.6," only to certain entities approved by the U.S. government before its official launch.
CEO Hassabis is said to have held prior discussions for months with the Donald Trump administration, European officials, and peers in the AI field before unveiling this proposal. President Trump signed an executive order early last month requiring AI corporations to submit their latest high-performance models to the government 30 days before release for security vetting.