Global cooperation needed to tackle AI threats, says Bank of England governor

Andrew Bailey, the governor of the Bank of England, has urged global cooperation to counter growing threats from artificial intelligence, cautioning that the US and the Trump administration cannot succeed on their own. His comments come weeks after President Donald Trump temporarily barred foreigners from using Anthropic's powerful Claude Mythos model.

Speaking to the Guardian ahead of his speech at the annual Mansion House dinner in London, Bailey said governments worldwide must join forces to prevent bad actors from obtaining powerful and potentially destabilising digital tools. He argued for stronger coordinated testing to ensure AI models are safe for wider deployment.

The Trump administration has already frustrated allies seeking a joint approach by temporarily banning Anthropic from allowing foreign users access to some of its most advanced models, including Claude Mythos, which experts warn could threaten cyber defences. Though the ban was lifted after a few weeks, Bailey said the US should recognise it may not be able to secure itself against rising cyber threats or establish robust recovery plans without global collaboration.

In his speech to City bosses on Tuesday, he elaborated: "No country can seal itself off from the cross-border nature of systems that are prevalent today." His call for action on AI coincided with a defiant Mansion House address by Chancellor Rachel Reeves, who defended her record and warned her successor not to squander the "platform of stability" she had created. Reeves cited falling borrowing, rising investment, productivity and wages, and a drop in waiting lists. However, she cautioned that the UK economy would face continued shocks, such as renewed Middle East hostilities.

Bailey emphasised that international cooperation is essential for addressing AI threats, particularly through coordinated testing. His remarks align with those of British Nobel laureate Demis Hassabis, who has called for a US-led global AI watchdog to test advanced models and hit the brakes on their development if risks become too high.