The BoE raised this week the potential need for an AI “kill switch”. The BoE is concerned about advanced AIs that drift from public policy objectives or amplify market volatility, particularly where multiple agentic AIs respond with a herd mentality to the same prompts and triggers. To protect against AIs that could cause serious market harm, the BoE is considering whether it should require organisations to build in AI kill switches. This has arisen shortly after the US exercised a regulatory kill switch on Claude, cutting off the UK from the most advanced Claude AI models. The risks posed by kill switches and how organisations may address them are discussed in our recent article.
BoE suggests AI “kill switch” needed
