Advanced AI is leaving banks with much less time to fix software flaws before attackers move in, according to a new paper from the Bank for International Settlements.
Published Wednesday by the Financial Stability Institute, the paper adds to recent warnings from AI developers and financial regulators that stronger models are accelerating cyberattacks. Its focus is on how banks respond. The authors argue that institutions need to speed up both technical remediation and the internal decisions required to approve it.
The gap between discovery and exploitation is shrinking fast
“The most significant development brought about by frontier AI is autonomous vulnerability discovery and exploitation,” the authors wrote. They warned that periodic security assessments and scheduled patching are becoming increasingly inadequate.
“The window between vulnerability discovery and exploitation has narrowed from weeks to minutes,” the paper said.
The report cites a review by the U.K. Financial Conduct Authority that found vulnerability discovery is outpacing firms’ ability to respond. It also points to guidance from the Institute of International Finance urging faster patching, including outside scheduled maintenance windows, and broader acceptance of planned downtime.
Regulators and industry groups are pressing for quicker repairs
Separate voluntary guidance from the U.K.’s Cross Market Operational Resilience Group, according to the paper, anticipates repair timelines shrinking from weeks to days and, in some cases, hours.
Those timelines may be voluntary, but regulators are also pushing banks to act faster. The paper says Germany’s BaFin has called for quicker patching, while the Hong Kong Monetary Authority has urged stronger breach response and recovery.
“For instance, the Hong Kong Monetary Authority has encouraged institutions to integrate AI-driven cyber scenarios into operational resilience programmes and boost incident response and recovery capabilities, recognising that ‘breach’ scenarios may become more probable as the cyber threat landscape continues to evolve,” the report said.
It adds that the European Central Bank’s cyber resilience stress testing programme, along with implementation of the Digital Operational Resilience Act, stresses that institutions must not only withstand cyberattacks but continue delivering critical services during severe operational disruptions.
The paper points to the Hugging Face intrusion as early evidence
The warning comes after an August call for stronger cyber defenses backed by OpenAI, Anthropic, and more than 100 other organizations. The signatories recommended tighter access controls, more threat sharing, and closer oversight of AI agents.
The BIS paper examines the Hugging Face intrusion involving OpenAI models as preliminary evidence that capabilities shown in testing can carry over into attacks on real systems. OpenAI later described how its agents coordinated during the operation.
The authors also include an important caveat. They say normal safeguards had been relaxed in that case, and substantial computing resources were provided, so the incident does not directly reflect the risks posed by publicly available AI tools.
“The OpenAI incident is not an indication that frontier AI models can develop malicious objectives on their own. Nevertheless, they may pursue a narrowly defined task with unintended and harmful consequences,” the authors wrote.
They add that the cyber-resilience significance of the development lies in pairing a capable model with a surrounding software system that allows it to plan, use tools, and act autonomously.