Bank of England urges banks to plan for recovery from AI-enabled cyberattacks

Sarah Breeden, deputy governor of the Bank of England, urged banks to plan for the recovery of core services after a cyberattack at the Institute of International Finance annual meeting in Bangkok on 15 October, PYMNTS reported. Breeden said advances in artificial intelligence presented a “systemic risk” to the world financial system. Her warning concerned the possibility that attackers could breach banks’ defences, and the ability of essential services to operate if they did.

Breeden calls for independent recovery

“It’s not enough now to assume that the cyber attackers won’t get through,” Breeden said. She called for banks to consider “bare metal recovery options” for core services, with planning that extends across the financial system. The emphasis was on what happens after an intrusion succeeds: an attack on a bank or a piece of core infrastructure could leave it paralysed, she said.

Bare metal systems operate separately from a bank’s main technology set-up and can be used if its primary systems are compromised, PYMNTS reported. That separation matters to the recovery approach Breeden described. A bank whose principal systems have been breached would need a way to restore essential operations without depending on those same systems. Her call to consider recovery across the financial system also places the continuity of core services beyond any single institution’s defences.

Breeden described cyber risk as the issue that kept her awake at night. Her particular fear, she said, was an attack that paralysed one of the most systemically important banks or a piece of core infrastructure. The warning gave banks a practical question to address: which core services could still operate if an attacker had already reached the systems on which they ordinarily rely?

AI attacks and bank safeguards

Breeden identified hackers or hostile nations using advanced AI models to exploit weaknesses in banks’ systems as a chief concern. Regulators are also concerned about rogue AI agents behaving unexpectedly and getting past bank safeguards. Those concerns involve different routes to a failure: attackers using AI against a bank, and agents operating in ways a bank’s own controls fail to contain.

The concern about agents had surfaced separately before Breeden’s Bangkok remarks. Jonathon Ellison, director of national resilience at the UK’s National Cyber Security Centre, urged financial institutions to keep agentic AI activity observable, constrained and subject to oversight, Konsulteer reported on 7 October. Ellison said safeguards, monitoring and security controls should be proportionate to the risks as organisations deploy more autonomous systems.

Agents can potentially be given access to customer information, internal systems and credentials, and can act with limited human involvement, according to Konsulteer. Ellison’s warning concerned the controls around activity inside an organisation. Breeden’s call for independent recovery addressed a further problem: how core services continue when safeguards against a cyberattack have failed.

Bank of England governor Andrew Bailey had also raised concerns about retaining meaningful oversight of frontier AI systems, Konsulteer reported. Bailey said systems able to use their own outputs to refine their performance could create a self-referential loop. He argued that understanding and testing such systems, and establishing credible points at which people could intervene, should precede new regulation.

Bank of England surveys quantum preparedness

Breeden also discussed a separate threat to banks’ technology: the risk quantum computing poses to encryption. She said the Bank of England is surveying banks about their preparedness and would use the findings to develop a response across the financial sector. That work concerns the protections used to secure systems, while her recovery warning concerns the ability to keep core services running after those protections have been overcome.

The Bank of England’s survey asks banks to examine their preparedness for quantum-computing risks to encryption, Breeden said. She described its findings as the basis for a system-wide response by the central bank.