British Prime Minister Rishi Sunak revealed plans on Thursday to establish an AI Safety Institute, marking a pivotal step in global AI governance. In a speech ahead of the UK's global

AI summit, Sunak outlined his vision for the institute, which he described as the first of its kind globally. He emphasized that the AI Safety Institute would significantly advance our understanding of AI safety by meticulously examining and testing new AI models to assess their potential risks, from social issues like bias and misinformation to more extreme threats. The Institute's findings will be shared with the world.

The institute's work will build upon the achievements of the UK's existing Frontier AI Taskforce, which has been granted privileged access to technology models from leading AI companies such as Google DeepMind, Anthropic, and OpenAI. While some information will be made public, certain sensitive national security information will be shared with a select group of like-minded governments, according to an unnamed UK official closely involved in summit discussions. The primary goal is to ensure model safety before their deployment.

Sunak also proposed the establishment of a global AI research network akin to the Intergovernmental Panel on Climate Change (IPCC) to release a comprehensive AI science report. The network, composed of experts nominated by participating countries and organizations, would be responsible for publishing this report. The UK's approach would set this new body apart from existing AI research networks, like the Franco-Canadian-led Global Partnership on Artificial Intelligence, by including Chinese representation.

Additionally, Sunak emphasized that the UK would make a concerted effort to achieve international consensus on the nature of AI risks and defend the inclusion of Chinese participation. He argued that an effective AI strategy requires engagement with all the world's leading AI powers. Photo by mikemacmarketing, Wikimedia commons.