The Bank for International Settlements said in a new report that advanced AI tools have reduced the time banks have to repair software vulnerabilities before attackers exploit them to just minutes.

The BIS said the main change introduced by cutting-edge AI is the autonomous discovery and exploitation of vulnerabilities. In its view, that shift sharply compresses the response window for financial institutions once a software flaw exists.

The report argues that traditional defenses are becoming less effective against that pace. Periodic security assessments and scheduled patch updates are no longer sufficient when attackers can identify and act on vulnerabilities far more quickly than before.

According to the BIS, banks need to accelerate software repair processes and speed up the related authorization decisions that often slow deployment. The report frames the issue as an operational challenge for institutions that rely on layered approvals before changes can be made to live systems.

The warning comes as AI adoption expands across finance, including in development, automation, and internal risk functions. That broader rollout has also increased attention on whether the same tools can strengthen cyber defenses fast enough to offset more capable offensive uses.

Why It Matters

The BIS warning matters because it shifts the discussion from general AI risk to a specific financial-stability issue: the shrinking time between a vulnerability appearing and an attack being launched. For banks and other financial platforms, resilience may depend less on scheduled controls and more on whether core systems can support near-real-time response.

That has broader implications for market infrastructure linked to digital assets as well. Crypto venues, custodians, and payment systems operate in an environment where software changes, integrations, and external dependencies are constant. If AI compresses the exploit window across finance, cyber readiness becomes more central to institutional trust in digital financial infrastructure.