Bailey urges global cooperation to tackle AI threats
TL;DR:
- Bank of England governor Andrew Bailey has called for international cooperation on frontier AI, warning the US cannot achieve its security aims alone.
- His comments follow the Trump administration’s temporary ban on foreigners using Anthropic’s Claude Mythos model.
- Bailey wants stronger coordinated testing before powerful models enter wider circulation.
Andrew Bailey has taken the Bank of England’s AI concerns beyond financial stability and into foreign policy, calling for international cooperation to tackle AI threats and warning that the United States cannot achieve what it wants on its own. Speaking to the Guardian before his Mansion House speech, the governor said governments must join forces to keep powerful and potentially destabilising tools away from bad actors.
An interconnected system cannot be sealed off
The argument is structural rather than diplomatic. “The US can’t achieve what it sensibly wants to achieve, in terms of strengthening defences, on its own because it is a highly interconnected system,” Bailey said, adding that better international understanding of how frontier AI is introduced would require stronger coordinated testing before models reach wider circulation. He put it more bluntly to City bosses: “No country can seal itself off from the cross-border nature of systems that are prevalent today.”
The context is a live grievance. Weeks earlier the Trump administration temporarily banned foreigners from using Anthropic’s Claude Mythos model, which experts warn could threaten cyber defences. The ban was lifted within weeks, but the episode left allies wary — and its consequences are still visible in Britain, where the government’s banking adviser says UK lenders remain unable to access the model.
Bailey’s intervention marks a clear escalation. This is the same governor whose Financial Policy Committee named AI a financial stability risk barely a week ago; he is now arguing the problem cannot be solved inside any single jurisdiction. The Bank is becoming a recurring voice in AI governance — a notable drift for an institution whose remit is monetary and financial stability, and a signal of how far AI risk has travelled from the technology pages into central banking.
The call for coordination also sits awkwardly beside a rival proposal made the same day. Demis Hassabis, the British Nobel laureate behind Google DeepMind, used an essay on X to call for a US-led standards body to test advanced models — leadership located in Washington rather than shared internationally.
Looking forward
Bailey’s comments landed alongside Rachel Reeves’ final Mansion House speech, in which she defended her record and warned her successor not to squander the stability she had built. With Andy Burnham due to take over as prime minister next week and a new chancellor expected, Britain’s AI governance voice may increasingly come from Threadneedle Street rather than the Treasury.