TL;DR
Andrew Bailey, the Bank of England governor, said on 9 November that artificial intelligence poses two distinct threats to financial stability: a sharper cyber threat from frontier models, and the chance that an AI share bubble bursts. Speaking at a University of Zurich lecture series run by the Swiss Institute of International Studies, he asked for wider testing of AI models and stronger reporting of incidents, while describing himself as an optimist on the technology.
Two risks, one speech
On cyber, Bailey was direct. “In our world, frontier AI has increased the risk from cyber-attacks – something that was already prominent,” he said, adding that the resulting concerns “may call into question public trust in AI. This is critical.”
He wants “more extensive model testing” both before and after models are released, and “better incident reporting, monitoring and follow up” for AI tools.
On markets, he set out how a correction could happen. Investors might be overvaluing AI companies, or the returns they expect could take longer to arrive. “There are good reasons to believe that the world moves faster now, but still the timing may not live up to market expectations,” he said. Either could trigger the bubble bursting, the governor said, with knock-on risk to UK financial stability.
The optimist’s caveat
Bailey balanced the warnings with a positive view of AI’s economic potential. AI, digital technology and “most probably robotics too”, he said, “have the potential to lift the sustainable growth rates of many economies through raising productivity growth”. He added: “That is a good thing.”
What it means
The speech pulls together threads the UK has heard from several institutions in recent days. A few days earlier, the NCSC, in its annual review, warned that AI-boosted attacks on critical national infrastructure could come within two years, and we also covered the Bank’s latest findings on AI and hiring.
In our view, the incident-reporting call is the line for UK firms to note. A central bank governor asking for “better incident reporting” on AI tools signals where supervisory expectations could head for banks and insurers that deploy them. Australia this week went further: after an OpenAI agent breached a government portal, its government said AI companies will have to report misaligned behaviour and cyber incidents, under AI legislation it plans to introduce next year. For financial services firms, keeping a log of AI incidents and model changes now is cheap preparation for whatever form UK expectations take.