Skip to news
← News
Analysis

Bank of England chooses guidance over new cyber rules

Follow us on Telegram

The Bank of England and its fellow regulators have made a deliberate bet: clearer expectations, rather than new rules, will make Britain's banks and insurers harder to break. The joint proposal published on 19 November 2026 treats cyber attack as a top threat to financial stability, yet leaves firms free to choose how they meet the standard. That choice now faces its sharpest test, because reported attacks have risen to near-record levels while artificial intelligence widens the field of attack.

Official document
DocumentDetails
Issued byBank of England
Published19 November 2026
Official textbankofengland.co.uk
Linked official pages2
Earlier documents compared1
Source: Bank of England.

Can guidance change behaviour when the danger is already inside the system? The proposal answers by gathering existing duties into one place and spelling out what good looks like across governance, defence and recovery. Its success will turn less on the wording than on whether boards, technology teams and suppliers act on it before the next serious incident, and whether the consultation that runs to 26 February 2027 produces a final version firms actually use.

Guidance without new rules

The Bank, the Financial Conduct Authority, known as the FCA, and the Prudential Regulation Authority, known as the PRA, issued the proposal together. It gathers existing duties and lessons from supervision into a single statement of what firms should do, and it aligns that statement with the Cyber Assessment Framework of the National Cyber Security Centre, the government's technical authority on cyber security. The regulators describe the result as guidance, not a fresh set of binding rules, so firms can choose the methods that fit their systems while meeting the same expectations.

That scope runs through six linked tasks. Boards and senior managers must own cyber risk, set a clear appetite for it and receive regular assurance. Firms must keep current maps of the technology, data and outside dependencies that matter most, and judge the threats and weaknesses that could affect them. Protection rests on sound configuration, prompt patching, control of who can enter systems, protection of data and oversight of suppliers. Detection requires monitoring, clear escalation and repeated testing of the ability to spot an attack. Recovery requires practised plans, clear decisions and communication, the ability to contain an incident and restore services, and a habit of learning from each event.

Attacks climb while confidence holds

The case for action rests on two measures that point in different directions. The Bank's survey of market participants, conducted between 27 July and 24 August 2026, found confidence in the stability of the financial system broadly unchanged from the first half of the year, even as respondents put the chance of a high-impact event higher. Fifty-seven firms took part, a response rate of 66%. Geopolitical risk and cyber attack remained the two risks named most often, and respondents also rated them the hardest to manage and the most likely to happen.

From the document
Original text, quoted as published.

Cyber resilience guidance consultation and AI and quantum effective practices News // Minutes 19 November 2026 Minutes of the Market Participants Group...

Source: Bank of England, 19 November 2026. Bank of England, Cyber resilience guidance consultation and AI and quantum effective practices

Reported incidents tell the sharper story. In October 2026, reports to the FCA reached their second-highest monthly total in 5 years. That total stood 52% above the earlier peak over 18 months, recorded in May 2025. Nikhil Rathi, chief executive of the FCA, said: "Technology moves fast, and so do cyber threats." His point, in the regulators' framing, is that speed now matters on both sides: attackers adapt quickly, and defences that wait for a new rulebook will lag behind them.

Artificial intelligence moves from edge to centre

Artificial intelligence has moved from a peripheral worry to a central one in the same survey. The number of participants naming risks around artificial intelligence rose sharply, reaching its highest recorded level in each of the three ways the survey asks about risk: as a source of danger, as a challenge to manage and as a risk likely to occur. A separate Bank and FCA survey found 88% of respondents expect frontier AI cyber capabilities to affect their risk profile moderately or highly over the next 2 years.

The regulators' separate paper on artificial intelligence treats the technology as both weapon and shield. It sets out practices firms can adopt to defend themselves with artificial intelligence, while warning against over-reliance on it and insisting that human oversight remains essential, especially for key decisions. In the same survey, 64% said awareness of frontier AI had pushed vulnerability management and patching up their agenda more than any other area. The paper is complementary to the main guidance and separate from it: its practices are not part of the regulators' formal expectations, and the regulators use the Organisation for Economic Co-operation and Development definition of an artificial intelligence system rather than looser labels that shift with fashion.

Quantum risk is distant and structural

Quantum computing presents a different kind of problem: not a current wave of incidents, but a future break in the foundations. The regulators encourage firms to consider its implications, because it could undermine the public-key cryptography that underpins financial transactions. That cryptography uses paired public and private keys to keep payments and messages secret and to prove who sent them; if a powerful quantum computer could break it, much of the trust built into digital finance would need rebuilding.

The quantum paper, like the artificial intelligence paper, is practical rather than regulatory. It offers effective practices for firms to prepare, without opening a formal consultation on the material itself. The regulators say they want to know whether this kind of support is useful and what industry would like to see in future. That invitation matters because preparation would be long and expensive, and firms will not commit at scale to a timetable nobody can yet fix.

What the consultation will test

The proposed guidance applies to firms already covered by the regulators' existing operational resilience requirements, the rules that require important services to keep running through disruption. The consultation closes on 26 February 2027, and feedback will inform the final guidance next year. Firms, trade associations, technology providers and other interested stakeholders are encouraged to respond.

The test, then, is not whether the document is well written. It is whether guidance can move patching, inventories, detection and recovery plans faster than the incident count is moving, and whether firms treat the artificial intelligence and quantum papers as early warning rather than optional reading. If responses show the expectations are clear enough to act on, the regulators will have avoided new rules without weakening the standard. If they do not, the next serious incident will reopen the argument for compulsion.

Source: Bank of England, Bank of England, latest Systemic Risk Survey, Bank of England, Effective practices: Cyber risk and resilience management relating to artificial intelligence (AI), Bank of England, PRA consults on updated guidance for friendly society amalgamations and transfers (22 July 2026) · Date:

Bank of England chooses guidance over new cyber rules

Read next