DeepMind CEO Calls for Global AI Watchdog: Humanity Has 'Precious Window' to Ensure AGI Safety

As Google DeepMind's CEO, Demis Hassabis is one of the most influential figures in the global AI field. The team he led developed milestone AI systems like AlphaGo and AlphaFold, earning extremely high reputation in both scientific and industrial circles. His public call for establishing a global AI regulatory body is considered an important statement on safety issues from an AI industry leader. In his essay, Hassabis pointed out that the current pace of AI technology development far exceeds the speed of regulatory framework construction, and failure to take timely action could lead to irreversible consequences. He emphasized that this is not just a technical issue but a major decision concerning humanity's future.

The core content of Hassabis's proposal is to establish a US-led international coalition specifically responsible for developing and enforcing safety standards for frontier AI models. This coalition would play a role similar to the International Atomic Energy Agency (IAEA), supervising, evaluating, and certifying AI systems developed by various countries. Specifically, this body would be responsible for several aspects: first, developing safety evaluation standards for AI models, including capability testing, risk assessment, and compliance checks; second, conducting mandatory reviews of AI models exceeding specific capability thresholds; third, establishing reporting and investigation mechanisms for AI incidents; fourth, promoting cooperation and information sharing among countries in AI safety research.

The backdrop to this proposal is that the AI industry is currently in a period of rapid development, with major companies continuously releasing increasingly powerful models. OpenAI's GPT series, Anthropic's Claude series, and Google's Gemini series are all constantly breaking performance records. However, as model capabilities increase, potential safety risks are also growing. For example, powerful AI systems could be used for malicious purposes such as cyberattacks, biological weapon design, and disinformation campaigns. Additionally, the emergence of AGI may bring deeper challenges, including large-scale transformation of employment markets, adjustment of social structures, and even redefinition of human identity. Hassabis believes that effective regulatory frameworks must be established before these issues become uncontrollable.

However, establishing a global AI regulatory body faces numerous challenges. First is the sovereignty issue — countries may be unwilling to hand over AI regulatory authority to international bodies. Second is the enforcement problem — ensuring countries comply with unified standards is difficult. Third is technical complexity — evaluating and regulating AI systems requires highly specialized technical capabilities. Fourth is the balance issue — excessive regulation may stifle innovation, while insufficient regulation cannot effectively control risks. Hassabis acknowledged these challenges in his essay but believes that compared to the risks of taking no action, these challenges can be overcome. He called on governments, tech companies, and academic communities worldwide to work together to establish an effective global governance framework before AGI arrives.