Google wants ethics committee first, 12 years later Hassabis wants to add the last barrier to AGI According to Beating monitoring, when DeepMind was sold to Google in 2014, one of Hasabisti's conditions was the establishment of an independent AI ethics committee. 12 years later, he is still pondering the same thing, but this time he wants to manage not DeepMind, but the entire cutting-edge AI industry. The Wall Street Journal revealed that before stepping down as CEO of DeepMind, Hassabis had already been pushing for a new regulatory mechanism with senior US government officials and other AI lab leaders. US Treasury Secretary Vincent and Presidential Science Advisor Michael Kratsios have both heard of his proposal. Bessent himself was also involved in the development of the Trump administration's AI strategy. Hassabis compares future institutions to an AI version of the International Atomic Energy Agency. Subtly, the US government has already done half of it. The White House launched a voluntary review framework for cutting-edge models in June, and the plan unveiled by Hassabis in July is very similar to its operational details: both first identify the cutting-edge models that need to be reviewed, both allow up to 30 days before release for review, and also involve the government in testing. Disagreement only exists at the last gate. The White House has explicitly stated that this mechanism cannot become a mandatory licensing or approval process; Hassabis wanted to continue pushing forward. Voluntary first, and once the evaluation mechanism is implemented, it will become an entry threshold. If the cutting-edge model fails to pass, it cannot be deployed in the United States. The testing has also expanded from network attacks to biological risks, agent bypassing security measures, and deceptive behavior. If necessary, it can even coordinate with several top laboratories to slow down together. [Original link]