Hassabis Advocates for AI Ethics Committee and Regulatory Framework

Hassabis proposed the establishment of an independent AI ethics committee as a condition for selling DeepMind to Google in 2014. Twelve years later, he plans to create a new regulatory framework for the entire cutting-edge AI industry. Before stepping down as CEO of DeepMind, Hassabis discussed this proposal with senior U.S. government officials and leaders of other AI laboratories, including U.S. Treasury Secretary Janet Yellen and President Biden's technology advisor Michael Kratsios. Hassabis likened the future institution to an AI version of the "International Atomic Energy Agency."

In June, the White House launched a voluntary review framework for cutting-edge models, and Hassabis's publicly announced plan in July is similar, outlining key models that require focused review, allowing submissions for review up to 30 days before release, and involving the government in testing. The difference lies in the White House's stipulation that this mechanism cannot become a mandatory licensing or approval process, while Hassabis hopes to set entry thresholds after the evaluation mechanism matures on a voluntary basis, with cutting-edge models that do not pass unable to be deployed in the U.S. Additionally, the testing scope will expand to include cyber attacks, biological risks, agent circumvention of security measures, and deceptive behaviors, with coordination of top laboratories to slow down if necessary.