Lord Tarassenko (CB)

My Lords, I too congratulate the noble Lord, Lord Holmes of Richmond, on his commitment to AI regulation and on securing today’s QSD. In the time available, I will follow the noble Lord, Lord Harper, and concentrate on AISI, the AI Security Institute. When Claude Mythos Preview was released by Anthropic to a select group of launch partners a few weeks ago, AISI was the only non-American government organisation to receive access to the model for safety testing. However, should the UK Government in three years’ time have the same view of AI governance as the current US President, AISI’s future would be under threat. The Government could decide overnight to repurpose it or even shut it down altogether. I therefore proposed a Private Member’s Bill, not drawn in the ballot, which would establish AISI as an independent statutory body. It would make provision for its functions in relation to the testing and evaluation of advanced AI systems, including pre-deployment and post-deployment assessment of risks, and for the institute to give advice to the Government and Parliament. Putting AISI on a statutory footing, beyond its current status as a unit within DSIT, would signal to the international community that the UK’s commitment to AI safety was permanent. We cannot wait: Anthropic is predicting that there is a greater than 50% probability that AI self-recursive improvement will be achieved by 2028. If AISI has become an independent statutory body by then, it will be able to provide regulatory advice on whether AI designed by AI is safe to use or whether it should be banned in the UK. Expert-driven regulation based on advice from an independent AISI, using Henry VIII powers if required, is the best option for AI safety in the UK. I urge the Minister and the Government to allocate some bandwidth to thinking about this critical issue.