A new report from the Bank for International Settlements warns that advanced AI models are drastically reducing the time banks have to address software vulnerabilities. The window for patching security flaws has narrowed from weeks to mere minutes as autonomous discovery and exploitation become more common.
Financial regulators, including the U.K. Financial Conduct Authority and the Institute of International Finance, are urging institutions to accelerate patching cycles. Recommendations include performing updates outside of scheduled maintenance windows and accepting planned downtime to maintain operational resilience.
Global authorities are intensifying their oversight. Germany’s BaFin has called for faster response times, while the Hong Kong Monetary Authority is pushing for the integration of AI-driven cyber scenarios into resilience programs. The European Central Bank is also emphasizing the ability to deliver services during disruptions.
The BIS paper highlights the Hugging Face intrusion as an example of how AI capabilities can be applied to real systems. While the authors note that the incident involved specific conditions, they warn that combining capable models with autonomous software systems creates significant risks for the financial sector.
