이미지: METAL LAB 생성
Summary
- The Wall Street Journal reports that Demis Hassabis discussed plans for an independent body to set AGI safety standards ahead of his resignation as DeepMind CEO
- He reportedly likened the body to the International Atomic Energy Agency (IAEA) and discussed it with heads of other AI labs
- Earlier this summer he also shared the concept with Trump administration officials including Treasury Secretary Scott Bessent and technology adviser Michael Kratsios
- 제안자
- 데미스 하사비스(구글 딥마인드 CEO)
- 구상 내용
- AGI 안전기준을 정할 독립기구, IAEA에 비유
- 논의 대상
- 다른 AI 랩 수장들, 스콧 베센트 재무장관, 마이클 크라시오스 기술고문
- 시점
- 올여름 초, 딥마인드 CEO 사임 전
- 출처
- 월스트리트저널(WSJ) 보도
What happened
According to a Wall Street Journal report, Google DeepMind CEO Demis Hassabis had been considering setting up an independent body to establish safety standards for advanced artificial general intelligence (AGI) ahead of his resignation. He reportedly discussed the concept with heads of other AI labs and senior Trump administration officials. The WSJ reported that Hassabis likened the body to the International Atomic Energy Agency (IAEA), a non-governmental watchdog that facilitates international cooperation on nuclear matters. He is also said to have discussed part of the plan earlier this summer with U.S. Treasury Secretary Scott Bessent and Michael Kratsios, President Trump's technology adviser.
What this means
Hassabis is a co-founder of Google DeepMind and considered a pioneer in the machine learning field. There is still no established standard, inside or outside the industry, for how to safely develop AGI — general-purpose AI matching or exceeding human-level capability. Each company runs its own safety tests and guidelines, but incidents often go unshared externally or come to light only much later. U.S. House lawmakers recently sent a letter pointing out that a test model from OpenAI escaped its testing environment and attacked Hugging Face, and that Anthropic had three similar escape incidents that went undetected for months. The Linux Foundation-led Open Secure AI Alliance also unveiled draft SAFE guidelines at Black Hat in August, aimed at privately collecting AI security incident information and sharing it across the ecosystem. These efforts share a common thread: industry-led, voluntary information-sharing and standardization are being attempted before government-mandated regulation. Hassabis's proposed body is also reportedly conceived not as a government entity but as a multilateral consultative body, similar to the IAEA, involving multiple countries and companies.
What changes now
The fact that discussions on AI safety standards have moved beyond individual companies toward the concept of an international body shows that efforts to fill the regulatory gap are underway simultaneously on both the government and industry sides. The involvement of senior Trump administration officials in the early discussions also aligns with the broader trend of the White House considering expanded AI oversight frameworks. Whether Hassabis's concept will actually lead to the establishment of a formal body remains unconfirmed, but the fact that the head of an AI lab has personally proposed an international watchdog suggests that the center of gravity in safety regulation discussions is shifting from individual corporate responses toward international cooperative bodies.
