The Morning Note
Three central banks just priced AI debt as a systemic risk
Three central banks flagged AI-linked debt concentration within days of each other, as a major bank plans to double its AI budget and breaches at three Korean banks raise fears of AI-powered attacks.
Three warnings in one week
A Morgan Stanley estimate cited in the record of the Bank of England's Financial Policy Committee meeting puts global AI-related debt issuance at around $450 billion, already more than double the total issuance in all of 2025. Around the same time, the European Central Bank said AI-related borrowing already accounts for roughly a quarter of credit growth to firms, and the Reserve Bank of Australia warned that the AI boom's debt-financing cycle is becoming more opaque and circular.
Why it's concentrated
What worries each of them is concentration. The RBA pointed to chipmakers financing the very neocloud firms that buy their chips, a loop it called circular financing. Lagarde made a similar point: global equity valuations sit in a small number of AI firms that are borrowing fast, so a sharp reassessment of their prospects could spill over to euro area investors and the wider economy.
The attack surface grew too
The RBA's review called vulnerability to AI-driven cyberattacks a key concern, and three banks in Korea spent this week proving the point. KB Kookmin and Hana disclosed breaches on Friday, a day after Shinhan, Hana's after an external hacking agent's attempt to gain unauthorized access to its sales support system. Local media reported traces of a Chinese-language AI penetration-testing tool on a server believed to have been used in the Shinhan attack. That is an early sign that AI agents are now doing some of the probing themselves.
Banks are spending anyway
None of this is slowing the money going into AI on the bank side. Bank of America said its Erica assistant has handled 3.6 billion transactions and that without it the bank would need 11,000 more people to answer calls, and its CEO said the AI expense budget will double next year. That is the trade-off sitting under every one of these warnings: the operational case for AI inside a bank is strong enough that spending keeps rising even as the macro case gets shakier.
What to watch this week
For a credit or risk team, the read is not to panic but to check two things: how concentrated your own bank's AI-linked exposure is among counterparties, and how fast your AI governance process can catch an agent probing your systems before a breach like this week's does. The central banks have started measuring the first problem, and whether banks follow is the story to watch over the next few months.