Banking Watchdogs Ramp Up AI Scrutiny as Stability Fears Mount, Sources Say

3-line summary

• The US Federal Reserve, OCC and FDIC are questioning banks on their use of advanced AI models, the Bank of England is pushing large banks to review recovery plans, and the European Central Bank is requiring eurozone banks to submit AI-driven cyberthreat response plans, according to a Reuters report citing six anonymous sources; the agencies declined to comment.
• Key concerns cited include AI-driven cyberattacks, autonomous AI agents acting beyond their authority or outside human control, reliance on Chinese technology and external AI vendors, and AI adoption outpacing the build-out of controls; one European bank executive reportedly warned serious market damage could occur within six months.
• US regulators are questioning banks of all sizes on AI agent limits, behavior monitoring, approval processes, auditability, human oversight and kill-switch mechanisms without yet setting specific rules. The BoE is favoring industry collaboration over new rules, urging banks to build "bare metal" recovery systems fully separated from main systems and use frontier AI to find and patch vulnerabilities. The ECB's July letter required eurozone banks to submit AI-cyberthreat response plans, including backup and recovery measures and mapping of AI usage and vendor dependencies, by the end of October.

This summary is our own processing of the source article; see the source link below for full context.