▲ Demis Hassabis in conversation with Sam Altman
The head of Google's artificial intelligence division has stated that as Artificial General Intelligence (AGI) approaches, there is a need to establish a U.S.-led regulatory body.
Demis Hassabis, CEO of Google DeepMind and a Nobel laureate in Chemistry, predicted on the social media platform X (formerly Twitter) on July 14 (local time) that "we are only a few years away from the arrival of AGI."
"The scale of the impact this technology will have is unprecedented, likely 10 times larger and faster than the Industrial Revolution," Hassabis emphasized. "AGI is more comparable to the discovery of electricity or fire than to the internet or mobile technology."
He argued that because the uncertainties and risks posed by the advent of AGI are significant, a stance of "cautious optimism" must be maintained, and a dedicated regulatory body should be established.
Hassabis, who is British, suggested that the U.S. should lead this initiative, stating, "Given its economic and technological status, the U.S. is in a favorable position to take the first step in establishing such a framework."
He proposed the Financial Industry Regulatory Authority (FINRA)—a private organization that regulates Wall Street under the supervision of the Securities and Exchange Commission (SEC)—as a model.
He suggested that the new body should be responsible for evaluating AI models and collaborating with federal agencies to test them in areas related to security, such as cybersecurity and biological risks.
To achieve this, he proposed that cutting-edge AI labs initially submit their models to the standards body for review 30 days before release on a voluntary basis, with the process becoming mandatory once its effectiveness is proven.
He cited the advantages of this approach as focusing on technological progress while simultaneously supporting innovation and encouraging responsible behavior.
He also emphasized that if a U.S.-led organization is established, it could lead to the creation of international standards for AI.
Hassabis's proposal is interpreted as being linked to the U.S. Department of Commerce's recent issuance of export restrictions on Anthropic's AI model, "Claude Mythos," prohibiting services to foreign nationals for a period due to security concerns.
OpenAI also released its latest AI model, "GPT-5.6," only after a process of pre-releasing it to select institutions approved by the U.S. government.
It is reported that before making this proposal public, Hassabis held prior discussions with the Donald Trump administration, European officials, and colleagues in the AI field over several months.
President Trump signed an executive order early last month requiring AI companies to submit their latest high-performance models to the government for security verification 30 days before their release.
Hassabis is known as the "father of AlphaGo," the Go-playing AI that defeated Lee Sedol in 2016, and he was a co-recipient of the 2024 Nobel Prize in Chemistry for his work in developing AlphaFold, an AI program that predicts protein structures.
(Photo: AP, Yonhap News)
※
Copying, redistribution, and unauthorized use in AI training are strictly prohibited.