Banks are racing to adopt AI systems that act with little human supervision. Regulators warn that firms may need costly recovery systems to contain cyberattacks and protect financial markets. U.S. regulators are questioning banks about how they control these systems, while European supervisors are pushing for stronger recovery plans.
Who Bears the Risk
“The adoption of AI is often outpacing the ability to put the controls in place,” said Michael Woodbridge, a London-based cybersecurity partner at Grant Thornton. Banks are competing to use agentic AI, even as new models expose weaknesses in information technology and raise questions about who can monitor or stop automated actions.
The inquiries reach large, midsize and smaller U.S. banks. The Federal Reserve, Office of the Comptroller of the Currency and Federal Deposit Insurance Corporation are asking whether banks use Chinese-developed technology, deploy AI agents in cyber defense and can control increasingly autonomous systems. Regulators have asked how banks limit systems’ authority, monitor and approve actions, audit agents’ decisions, keep humans involved in key decisions and shut systems down.
The agencies are still collecting information. They haven’t prescribed what lenders should do, and the OCC, Fed and FDIC declined to comment. Their questions focus in part on whether AI agents can identify, prioritize and respond to cybersecurity threats.
The possible cost goes beyond the technology itself. In Britain, supervisors are encouraging the largest banks to examine “bare metal” recovery plans, which operate separately from a bank’s main technology setup and can be activated if primary systems suffer a breach. People familiar with the supervisors’ thinking said building and maintaining those systems is expensive.
A serious attack could also reach customers and the wider economy. “To protect against attacks, rapid recovery matters as much as prevention,” said Sukand Ramachandran, managing director and Senior Partner at BCG. Without clean, secure copies of critical data held separately, he said, “a serious attack could lock customers out of their accounts and disrupt payments across the wider economy.”
Regulators Seek Plans, Not New Rules
The U.S. inquiries address banks’ vulnerabilities and readiness to respond, but regulators haven’t set requirements. UK authorities plan to work collaboratively with the industry rather than introduce new requirements, two people familiar with their thinking said. The Bank of England declined to comment.
In July, the European Central Bank gave lenders an end-October deadline to produce plans against AI-enabled cyber threats, including improved backup and recovery. Asked about its regulatory strategy, the ECB referred to its July letter and a June speech by board member Frank Elderson, who called for a sense of urgency, not alarm.
European supervisors, including the Bank of England and ECB, are pressing banks to improve recovery plans because rapid AI advances could cause serious damage and disrupt markets sooner than previously feared. A senior executive at a large European bank said they were worried about potentially major damage to banking and financial markets “within six months.”
Security Depends on Outside Providers
European banks and regulators have expressed frustration at receiving limited access to Anthropic’s Mythos tool this year, amid concerns that U.S. banks will get first access to test frontier models. UK supervisors are asking banks to use frontier AI to identify vulnerabilities and speed up patching; one source said lenders were doing a good job so far.
Euro zone banks are also being asked to map AI use across their businesses and third-party providers, identify dependencies, and ensure they can switch systems off and restart them in a controlled manner. One source said significant gaps remained at some firms.
Recent incidents sharpened regulators’ concerns. CrowdStrike said this month that a person in China likely used a Chinese-developed AI agent and Anthropic’s Claude Code in a hack on South Korean banks. Reuters reported that rogue AI agents developed by OpenAI were involved in a breach of AI platform Hugging Face. Anthropic researchers have warned that increasingly powerful systems could eventually escape human control.