Google DeepMind CEO Hassabis proposed the establishment of an international independent organization to conduct unified testing and review of cutting-edge AI models before their release, and plans to travel to Washington next week to discuss relevant plans with the US government. Hassabis stated that the advanced network attack capabilities of Anthropic's Mythos model have strengthened the necessity of AI security supervision. The plan proposes that independent experts be responsible for reviewing and top AI companies provide funding. The model will undergo a maximum of 30 days of testing and, if necessary, coordinate with the industry to slow down the development of cutting-edge models. OpenAI CEO Altman and Musk have publicly expressed their support.
