Skip to content

UK edition London closed

Markets

  • FTSE 10010,346.34-1.02%
  • S&P 5006,910.39-3.70%
  • GBP/USD1.3082-0.49%
  • EUR/USD1.0975+0.03%
  • Brent$89.95-2.21%
  • Gold$3,515.40-4.62%
  • Bitcoin$57,789-1.72%

Delayed · 09:03 UK

UK regulators warn AI is shrinking the time firms have to patch flaws

The PRA, FCA and Bank of England say AI is compressing the gap between vulnerability discovery and attack, in non-binding practices outside the cyber consultation.

The PRA, FCA and Bank of England said on Thursday that AI is shortening the time between finding a software flaw and exploiting it, and that firms should move from periodic patching to exposure management.

The warning sits in non-binding “effective practices” on AI, published alongside a joint cyber consultation paper that closes on 26 February 2027. The AI material is not part of the regulators’ formal expectations and is not open for formal consultation.

The regulators said in the AI practices that recent developments are “having a material impact” on cyber risk by cutting the cost, skills and time threat actors need. Advanced systems can let attackers exploit vulnerabilities within hours or days rather than weeks or months.

Exposure management means identifying any misconfigurations, security gaps or weaknesses a threat actor could exploit, and managing them systematically. The regulators said traditional vulnerability management only identified and fixed flaws periodically. They now want continuous monitoring, starting with external-facing systems and working inwards.

They also said cyber risk is “no longer a purely technical issue” but a core business risk and leadership responsibility. Boards are expected to seek assurance that controls will work during incidents, not only that they are designed well.

Survey: 88% expect moderate or high impact

A Bank and FCA survey found 88% of respondents expect frontier AI cyber capabilities to have a moderate or high impact on their risk profile over the next two years.

In the same survey, 64% said vulnerability management and patching had seen the greatest increase in management attention because of awareness of frontier AI, according to the Bank’s news release.

Incident reports are rising. October 2026 had the second-highest monthly number of cyber incidents reported to the FCA in five years, 52% above the previous 18-month peak in May 2025.

Katharine Braddick, chief executive of the PRA and deputy governor at the Bank, said: “Cyber attacks are becoming more frequent, complex and dangerous.”

Nikhil Rathi, chief executive of the FCA, said: “Our response shouldn’t always be more rules.” He said firms had asked for clarity on what good cyber resilience looks like.

Hot patching and AI oversight: two cautions on 19 November

The regulators encourage automatic patching where feasible, particularly hot patching, meaning automatic patching without service disruption. They also said firms may use AI to strengthen exposure management, but that advanced systems are non-deterministic and can give different results across repeated assessments, so outputs need validation.

Access to advanced AI “does not automatically create or enhance effective cyber defence”, they said. Human oversight must be “meaningful, risk-based and capable of changing outcomes”.

The regulators added that AI-related cyber risks could threaten financial stability if they affect shared resources used by many firms, including critical third parties and open source software.

The practices are meant to be applied in proportion to a firm’s size, complexity and risk profile.

The consultation paper itself does not propose new rules but clarifies how firms should meet current requirements on governance, operational resilience, operational risk and third-party risk management in a cyber context.

The AI and quantum documents are complementary to, but separate from, the draft supervisory statement and final guidance.

Earlier warnings in May and June 2026

The Bank, FCA and HM Treasury said in a joint statement in May that the cyber capabilities of current frontier AI models were already exceeding what a skilled practitioner could achieve, at higher speed, greater scale and lower cost.

They said firms should consider automated and AI-enabled defences.

The Five Eyes cyber agencies, including the National Cyber Security Centre (NCSC), said in a statement on 22 June that delays in patching increase risk, and that the timeline for frontier AI disruption “is not years, it is months”. Richard Horne, NCSC chief executive, signed it for the UK.

On the separate consultation, Horne said the NCSC “welcomes this proposed guidance”, which he said would help firms strengthen resilience in line with its Cyber Assessment Framework.

Firms, trade associations and technology providers have until 26 February 2027 to respond to the consultation, and feedback will inform final guidance next year.

Published by Financial News