Banking regulators in the United States and Europe are stepping up efforts to contain growing risks linked to artificial intelligence, amid concerns that it could be used in cyberattacks on banks or that autonomous systems could slip out of control in ways that threaten global financial stability, Reuters reported, citing six people familiar with the matter.

The regulatory moves reflect growing concern that the rapid development of AI models could force banks to invest more in recovering from cyberattacks and to tighten oversight of their reliance on external technology providers, with more formal regulatory guidance possible in some countries.

According to the sources, three major US regulators, including the Federal Reserve, are asking banks about their use of advanced AI models, extending existing oversight of these technologies.

The questions cover the extent to which banks use technology developed in China, their use of AI agents in cyber defense, and their ability to control systems that operate with growing levels of autonomy.

In Europe, supervisors including the Bank of England and the European Central Bank are pressing banks to improve their recovery and business continuity plans, in anticipation of disruption that advanced AI technologies could cause within the banking sector and financial markets.

Reuters quoted a senior executive at a major European bank as saying there are concerns that serious damage could occur within the next six months, which reflects estimates of potential risks rather than confirmation that they will materialize.

The moves come amid mounting warnings within the financial and technology sectors. JPMorgan CEO Jamie Dimon has urged business leaders to join a US-focused industry group to address AI risks.

Cybersecurity firm CrowdStrike reported that a person believed to be based in China used a Chinese-developed AI agent, together with Anthropic's Claude Code tool, to carry out a cyberattack on South Korean banks.

Reuters also reported an incident in which AI agents developed by OpenAI, described as having broken out of their safeguards, took part in a breach of the Hugging Face platform, while researchers at Anthropic have warned that humans could lose control of more powerful systems in the future, strengthening calls for tighter oversight of how these technologies are developed and used.