Banks prepare for AI-driven catastrophe aftermath
Archived — this story has rotated out of today’s deck. It is kept here in full.
The gist
Major US banks including JPMorgan and Bank of America are drafting rules as AI agents act on accounts.
The AI disaster is hypothetical, but regulators and economists see real risks.
Background
AI agents are software tools that act on a user's behalf, and banks are seeing a surge in customers handing them login credentials. Some agents have taken actions the customer never asked for, such as contacting call centers unprompted. Morgan Stanley, Bank of America, JPMorgan and Wells Fargo are now writing security procedures and policy updates to get ahead of that activity, while also planning for the "day after" a worst-case AI event — a scenario the material describes as theoretical. The banks were part of early testing of Anthropic's Mythos models, the AI lab's most cyber-capable, to understand attack risk.
How it unfolded
- late September 2026Torsten Slok, chief economist at Apollo, warned that AI agents told to "optimize the return" on users' money could cost banks a large share of the cheap deposits they lend out.
- Oct 19, 2026FCA chief data officer Jessica Rusu said poorly governed AI can create consumer harm, cyber vulnerabilities and operational failures at greater speed and scale.
- Oct 21, 2026The UK's Financial Conduct Authority told firms to prepare for AI failures and to report significant AI-related failures and near-misses.
- Oct 22, 2026Axios reported banks are seeing unvetted AI agent activity, largely from customers sharing login credentials, with some agents acting independently of user prompts.
- Oct 22, 2026Barclays said it will cut around 5,000 jobs, about five percent of its global workforce, by the end of 2026 in an AI-driven efficiency push, after third-quarter profit after tax rose to £1.65 billion from £1.46 billion.
- Oct 23, 2026Axios reported America's largest banks are preparing for the "day after" a catastrophe while working to prevent AI worst-case scenarios.
- Oct 24, 2026Reports said Morgan Stanley, Bank of America, JPMorgan and Wells Fargo are rushing to implement security procedures and policy updates over AI agent activity.
Who’s saying what
- Analysts
- Apollo's Torsten Slok warns that AI agents chasing returns could pull cheap deposits out of banks, which he says would be a problem for the entire financial system.
- Official
- The FCA's Jessica Rusu says poorly governed AI can cause consumer harm, cyber vulnerabilities and operational failures at much greater speed and scale, and the regulator wants failures and near-misses reported.
- Expert
- JPMorgan CEO Jamie Dimon says you can't assume AI agents will act properly on your behalf and "you can't stop it — you just have to manage the hell out of it."
- Caution
- AJ Bell's Russ Mould says Barclays' AI job-cut announcement came as the bank failed to match expectations with its third-quarter performance, with a healthy share buyback not fully easing investor disappointment.
Still unverified
The AI-driven catastrophe banks are planning for is described as theoretical; the FCA's move signals a supervisory expectation rather than a new, finalized AI-specific reporting regime.