BoE, FCA and Treasury issue joint warning on frontier AI cyber risk

TL;DR:

  • The Bank of England, FCA and HM Treasury have published a joint statement telling regulated firms that frontier AI models already exceed what a skilled human attacker can do — and to plan and invest accordingly.
  • Five action areas are named: board-level governance, vulnerability identification, third-party and supply-chain risk, protective controls, and response and recovery.
  • The statement does not introduce new rules, but reinforces existing operational-resilience expectations and signals firms relying on end-of-life systems will be increasingly exposed.

UK financial regulators have closed ranks on frontier AI as a cyber-resilience problem. In a joint statement issued on 15 May, the Bank of England, FCA and HM Treasury said current frontier models “are already exceeding what a skilled practitioner could achieve, and at a significantly higher speed, greater scale, and lower cost,” warning that if used maliciously these capabilities “amplify cyber threats to firms’ safety and soundness, customers, market integrity, and financial stability.”

What firms are being told to do

The statement is not a new rule but a sharpening of existing operational-resilience expectations. It names five domains where the regulators expect active steps:

  • Governance and strategy — boards and senior management should understand frontier AI risk well enough to set strategic direction and approve investment, including insurance.
  • Vulnerability identification — firms must triage and remediate at much greater speed and scale, using automation where appropriate.
  • Third-party and supply-chain risk — open-source dependencies and external services need active monitoring and rapid patching when third parties surface vulnerabilities.
  • Protection — access management, network security and data protection should be designed to limit the attack surface a frontier model could reach, with AI-enabled defences considered to match attack speed.
  • Response and recovery — firms should be able to recover quickly from disruption and refer to the joint Bank-PRA-FCA cyber resilience practices published in October 2025.

The unambiguous regulator framing is consistent with BoE Governor Andrew Bailey’s April warning that Anthropic’s Mythos model could “crack the whole cyber risk world open.” Reuters confirmed on 15 May that the statement was issued jointly by the three authorities, and a separate Reuters report this morning revealed that Anthropic will brief the Financial Stability Board — chaired by Bailey — on the Mythos vulnerabilities directly.

Why this matters to UK SMEs in the supply chain

The statement targets PRA- and FCA-regulated firms, but the supply-chain language is the practical hook for UK SMEs. Any business providing software, libraries or services into UK banks, insurers or FMIs will now face hardened third-party-risk diligence — and firms running end-of-life systems or unpatched dependencies are being put on explicit notice. The Cross Market Operational Resilience Group (CMORG) and NCSC will be the channels for further guidance, including CMORG’s 14 May Frontier AI Risk Mitigation webinar.

Looking forward

The statement positions frontier-AI cyber risk alongside the existing operational-resilience regime rather than treating it as a standalone regulatory domain. That makes it harder for firms to defer — the expectations attach to rules already in force. Expect supervisory questions to escalate quickly, particularly on third-party and end-of-life-system exposure. CMORG and NCSC publications, including NCSC’s “vulnerability patch wave” preparation guidance, are the practical reading list firms should put in front of their boards before the next operational-resilience review.

Share this article