PRA CP17/26 Cyber Resilience: CAF Benchmark, AI and Quantum Practices
On 19 November 2026 the Prudential Regulation Authority (PRA), the Bank of England and the Financial Conduct Authority (FCA) published a joint consultation on cyber risk and resilience management: PRA CP17/26, issued by the FCA as GC26/9. It contains one draft document that would serve as a supervisory statement for the PRA and the Bank and as non-Handbook Finalised Guidance for the FCA (the draft SS/FG). The regulators say it adds no new rules, and they also say they will use it when they supervise and enforce the existing rules it lists, from the PRA’s Operational Resilience Parts to SYSC 15A of the FCA Handbook. Responses are due by Friday 26 February 2027.
The FCA’s announcement gives the backdrop: October 2026 saw the second-highest monthly number of cyber incidents reported to the FCA in five years, the highest in the past 18 months and 52% above the previous high recorded in May 2025. Alongside the draft, the three regulators published two effective-practices documents, one on artificial intelligence (AI) and one on quantum computing. Both are non-binding and sit outside the formal consultation, although the regulators invite comments on them.
For operational resilience, cyber and third-party risk teams, the working question is which existing artefacts the draft would be read against: the annual operational resilience self-assessment, board management information (MI), asset and third-party inventories, the cyber incident response and recovery plan, impact tolerances and the testing programme.
Related reading: SS2/21 Outsourcing: The PRA Register and Notification Guide
PRA CP17/26 and GC26/9: the dates that matter
- 19 November 2026: CP17/26 / GC26/9 published, with the AI and quantum effective-practices documents.
- Friday 26 February 2027: consultation closes. Responses go to CP17_26@bankofengland.co.uk and GC26-9@fca.org.uk. Feedback on the AI and quantum documents can be sent with the consultation response.
- 18 March 2027: the PRA’s operational incident and third-party reporting rules in PS7/26 apply, and the FCA’s SUP 15.18 (notification of operational incident), SUP 15.19 (notification of material third party arrangements) and SUP 16.33 (material third party arrangements register) come into force.
- First half of 2027: the FCA’s GC26/9 page gives this as the expected timing of the final guidance. The consultation paper says the regulators expect to publish a Policy/Feedback Statement containing the final SS/FG.
- 25 October 2027: qualifying cryptoasset firms join the scope list.
The consultation paper proposes no implementation period and no date from which the final SS/FG would apply. My reading is that this follows from its framing as clarification of rules already in force, so there is nothing in the text to plan a transition window around.
The March 2027 reporting date matters more than its position in the list suggests. Footnote 12 of the draft defines a major incident, for the SS/FG’s purposes, as any cyber incident reportable under the operational incident reporting requirements in the Notifications and Regulatory Reporting Parts of the PRA Rulebook and/or SUP 15.18, or one the regulators would expect notice of under PRA Fundamental Rule 7 or Principle 11 of the FCA’s Principles for Businesses. The Chapter 3 outcome draws on both the operational incident reporting requirements taking effect on 18 March 2027 and the broader expectation that firms notify their regulators of matters covered by PRA Fundamental Rule 7 or FCA Principle 11. Board escalation criteria under the proposed SS/FG should therefore not be limited to incidents that meet the new operational incident reporting thresholds.
How guidance with no new rules is meant to apply
Paragraph 1.3 of the consultation paper states that the proposals would not introduce new rules and instead clarify how firms should meet current requirements on governance, operational resilience, operational risk and third-party risk in a cyber context. Paragraph 2.1 of the draft SS/FG adds the operative sentence: the regulators have no specific rules for cyber risk and resilience management, but they will use the SS/FG when they supervise and enforce firms’ compliance with the rules listed in Annex 2 where those rules relate to cyber risk and resilience.
Annex 2 is therefore the map of where the expectations bite. Extracts by population:
| Population | Requirements, guidance and near-final provisions cited in Annex 2 (extracts) |
|---|---|
| Banks (PRA) | Fundamental Rules; General Organisational Requirements Part chapters 2 and 5; Operational Resilience Parts with SS1/21; Senior Management Functions Part with SS28/15; Internal Capital Adequacy Assessment Part rules 10.1 and 10.2; SS5/16; SS2/21; PS7/26 |
| Insurers (PRA) | Fundamental Rules; Conditions Governing Business Part chapters 2, 3 and 7; Insurance Operational Resilience Parts; Insurance Senior Management Functions Part; SS35/15; SS5/16; SS1/21; SS2/21; PS7/26 |
| FMIs (Bank) | FMI Fundamental Rules 5, 6, 7, 9 and 10; the Bank’s near-final UK Central Counterparties Instrument 2027 provisions on governance, business continuity and operational resilience, which had not yet been made or brought into force; UK CSDR articles on governance, operational risk and business continuity; Bank Codes of Practice on governance, outsourcing and operational resilience for RPSOs and SSPs; SS6/24 on critical third parties |
| FCA-regulated firms | PRIN; SYSC 2, 3, 4, 5, 7, 8, 9, 13, 14, 15A, 21 and 24; SUP 10C; SUP 15.18, 15.19 and 16.33 (from 18 March 2027); FG26/3, FG26/4 and FG16/5; Payment Services Regulations 2017 regulation 98; Electronic Money Regulations 2011 regulations 6, 10 and 11 |
Treating the absence of new rules as a reason to wait misreads that structure. Once final, the expectations would become the reference point for rules firms already have to meet, and paragraph 1.5 of the consultation paper says the regulators are likely to make additional regulatory and supervisory interventions in this area in the future.
The bank row includes rules 10.1 and 10.2 of the Internal Capital Adequacy Assessment Part. Annex 2 lists them without comment, and the draft does not explain what supervisors would test under them.
Scope of the draft SS/FG, and the firms left outside it
The scope follows the existing operational resilience perimeter. The consultation paper’s cost model counts 427 PRA-regulated firms, 10 Bank-regulated FMIs and around 1,700 FCA solo-regulated firms (footnote 5).
| Sector | In scope (draft SS/FG paragraph 1.9) | Outside scope (paragraphs 1.11 and 1.12) |
|---|---|---|
| Banking | UK banks, building societies, PRA-designated investment firms | Credit unions; UK branches of overseas banks |
| Insurance | UK Solvency II firms, the Society of Lloyd’s and its managing agents | Non-Solvency II insurers; UK branches of overseas insurers |
| FMIs | UK CCPs, UK CSDs, UK recognised payment system operators, UK specified service providers | Non-UK FMIs |
| FCA solo-regulated | Firms subject to SYSC 15A: enhanced scope SM&CR firms, UK recognised investment exchanges, e-money institutions, payment institutions, registered account information service providers, consolidated tape providers; qualifying cryptoasset firms from 25 October 2027 | FCA solo-regulated firms not subject to SYSC 15A |
The branch exclusion comes with a qualification. Paragraph 1.11 says the regulators may refer to the SS/FG’s expected outcomes when they assess a UK branch’s implementation of a cyber framework developed at parent or group level, and footnote 9 adds that paragraph 1.8 of SS44/15 also applies to third-country insurance branches. A UK branch whose cyber framework is set by a foreign parent remains outside the stated scope, but the regulators may refer to the proposed SS/FG outcomes when assessing how the branch implements that group-level framework.
Paragraph 1.12 addresses credit unions, non-Solvency II insurers and FCA solo-regulated firms not subject to SYSC 15A. It says those firms should manage cyber risk in line with relevant regulatory objectives, Fundamental Rules, Principles for Businesses and other requirements. They are encouraged to consider the SS/FG and NCSC guidance, and to assess whether Cyber Essentials and cyber insurance are appropriate. UK branches of overseas firms and non-UK FMIs are addressed separately in paragraph 1.11.
The NCSC Cyber Assessment Framework as the supervisory benchmark
Paragraph 1.14 of the draft asks every in-scope firm to refer to the outcomes, objectives and indicators of the National Cyber Security Centre’s Cyber Assessment Framework (CAF) to develop and assess the maturity of its cyber risk and resilience. It then states that the regulators will use the CAF, and future versions of it, as a key benchmark in supervisory engagement. Question 5 of the consultation asks whether firms agree.
Each technical chapter signposts the CAF principles it expects firms to use:
| Draft SS/FG chapter | CAF principles signposted |
|---|---|
| 3: Governance | A1 Governance, A2 Risk management, A3 Asset management |
| 4: Identification and assessment | A2 Risk management, A3 Asset management, A4 Supply chain |
| 5: Protection | B1 to B6: service protection policies, identity and access control, data security, system security, resilient networks and systems, staff awareness and training |
| 6: Detection | C1 Security monitoring, C2 Threat hunting |
| 7: Response and recovery | D1 Response and recovery planning |
| 8: Assurance and continuous improvement | A2 Risk management, D2 Lessons learned |
Internationally active firms can draw on their compliance with non-UK regimes as supporting evidence. Annex 2 lists DORA, 23 NYCRR Part 500, APRA CPS 230 and CPS 234, the MAS Technology Risk Management Guidelines and OSFI’s technology and cyber guideline among relevant non-UK publications, and the paper names the NIST Cyber Security Framework, the Cyber Risk Institute Profile and ISO standards as cross-sectoral references. The limit is stated plainly in paragraph 2.18 of the consultation paper: “Interoperability does not mean duplication, equivalence or homogeneity.” Under paragraph 2.6 of the draft, a firm remains responsible for demonstrating how it meets the UK expectations, including where its policies are set by a group outside the UK. A DORA control library is evidence to cite in that demonstration, and the firm still has to map it to the UK outcomes.
FMIs get a second benchmark. Paragraph 1.14 says Bank-supervised FMIs should consider the CPMI-IOSCO Principles for Financial Market Infrastructures and related guidance, which the Bank will use as a global benchmark. Footnote 16 of the draft goes further and points all firms to page 8 of the CPMI-IOSCO cyber resilience toolkit for example metrics on cyber risk and resilience, a document covered in our note on the CPMI-IOSCO cyber resilience toolkit for FMIs.
The phrase “and future versions thereof” means the benchmark is not frozen at the version a firm last assessed against. A CAF self-assessment that records which CAF version it used will be easier to defend when the NCSC revises the framework.
Board oversight, SMF24 and MI that separates planned from validated
Chapter 3 sets the governance outcome: the board approves and oversees the cyber risk and resilience strategy, holds senior managers accountable, allocates resources and, where appropriate, receives escalations of major cyber incidents and risks. Paragraph 3.7 adds that the board should not rely too much on a single specialist, and paragraph 3.8 says every board member needs enough knowledge to exercise oversight without becoming a cyber specialist. Where collective expertise falls short, paragraph 3.9 expects the board to be able to explain to the regulators how it will remediate, for instance through workshops, case studies of real incidents and participation in exercises.
For SM&CR firms, paragraph 3.19 says the Chief Operations Function (SMF24) should ordinarily have overall responsibility for implementing the cyber strategy and reporting on it to the board. The word “ordinarily” carries weight: a firm may allocate the responsibility to another SMF with appropriate authority, experience and functional independence, and the draft gives the Chief Risk Officer (SMF4) as an example. Firms outside SM&CR assign it to an appropriate senior manager, and FMIs allocate it under UK EMIR, UK CSDR, the Bank’s Codes of Practice and the FMI Fundamental Rules (paragraph 3.20).
The MI expectations are the most specific part of the chapter. Paragraph 3.16 lists, subject to proportionality:
- backlogs of assets with unpatched or critical vulnerabilities, assets nearing or beyond end-of-life or end-of-support, and out-of-policy credentials;
- detection performance, such as mean time to detect potential major incidents;
- major incidents and the time taken to resolve them, including the number or percentage resolved within recovery objectives such as impact tolerances, and recovery testing outcomes;
- privileged access information, including recently completed reviews;
- third-party data on concentration and on material third parties that have not provided appropriate cyber assurance;
- current and proposed AI use cases, with key dependencies and risks.
Paragraph 3.15 asks for MI that distinguishes current capabilities, planned improvements, and improvements already implemented and validated. Paragraph 3.14 asks it to highlight cyber risks the firm cannot currently manage within tolerance, with remediation timelines and/or compensating controls. The draft also supplies its own example of misleading MI in paragraph 3.17: a high percentage of patched vulnerabilities may not give an accurate or complete view of the firm’s overall cyber risk position.
Two smaller items sit in the same chapter. Paragraph 3.21 brings shadow ICT into each senior manager’s remit, including unauthorised use of AI and unauthorised development of software tools with generative AI. Paragraph 3.10 asks firms to consider the availability and suitability of cyber insurance and, where they hold it, to assess exclusions and limits on cover. Documentation of the strategy, policies and assurance programme should be updated as appropriate, for instance at the annual update of the operational resilience self-assessment, and provided to the regulators on request (paragraph 3.13).
Asset and risk identification built from existing registers
Chapter 4 expects an accurate, thorough and up-to-date inventory of every asset relevant to cyber resilience: data, facilities, hardware, telecommunication networks, people, services, software (proprietary, third-party and open source), supporting infrastructure such as energy supply, and third-party and key supply chain dependencies, including intra-group ones (paragraph 4.2). Paragraph 4.4 tells firms to build it from documents they already hold:
- asset inventories and bills of materials, including cryptographic bills of materials, AI inventories and hardware and software bills of materials;
- model inventories, with a cross-reference to Principle 1.2 of SS1/23 on model risk management;
- material third party arrangements registers;
- operational resilience maps and self-assessments.
The third item links this consultation to the reporting regime arriving on 18 March 2027, when the FCA’s SUP 16.33 register requirement comes into force and PS7/26 brings in the PRA’s third-party register requirements. A register built for submission can double as an input to the cyber inventory, provided the two are reconciled.
Paragraph 4.3 asks firms to identify concentrations and common dependencies that could cause correlated disruption across multiple important business services, group entities or recovery arrangements. On that wording, a backup or failover service that shares a dependency with the production service it protects falls inside the concentration analysis. Physical risks are in scope too: paragraph 4.9 names data centres, energy supply, hardware, telecommunications infrastructure and undersea cables.
On threat intelligence, paragraph 4.12 lists sources from incidents and near-misses at the firm, its material third parties and peers to the Government’s National Risk Register, NCSC advisories, specialist providers and vulnerability databases. Paragraph 4.13 sets the test supervisors would apply: a firm should be able to explain how its risk assessment informed its strategy and its decisions on resources, prioritisation and choice of controls. Paragraph 4.15 places a specific expectation on the board and senior management to set a risk appetite on key control indicators for vulnerability management, such as the frequency and scope of assessments.
Protection and detection: the controls the draft names
Paragraph 5.3 asks firms to assess their capabilities against published lessons and thematic findings from CBEST, STAR-FS, SIMEX and the Cyber and Operational Resilience Stress Test “irrespective of whether they took part in them”. The Bank’s 2025 CBEST thematic, drawn from 13 assessments in 2025 and dated 20 January 2026, recorded weak detection, which paragraph 3.30 of the consultation paper calls a common weakness among firms, alongside weak or poorly stored credentials, overly permissive access and unpatched systems. A firm that has never been through CBEST is still expected to read those findings against its own controls.
The protection expectations in Chapter 5 that change control design most directly:
- “Users” include AI agents and automated processes, so least privilege applies to them as it does to staff (paragraphs 5.4 and 5.5).
- Access policies and testing procedures should cover break-glass accounts for emergencies (paragraph 5.5).
- Multi-factor or other stronger authentication is expected where compromise could affect important business services or critical operations (paragraph 5.6), with access processes reflecting zero trust (paragraph 5.7).
- Cryptographic controls should follow established standards, including post-quantum cryptographic (PQC) standards where relevant, and firms should know in which jurisdictions third parties process their data (paragraph 5.11).
- Patching should use only authentic vendor-sourced and validated updates, with compensating controls for end-of-life assets and tested rollback procedures (paragraph 5.16).
- Where relevant, controls should cover customers’ end-to-end digital journeys, including APIs, authentication and consent (paragraph 5.20).
Chapter 6 expects detection in real time or near real time, prioritised on the assets, systems and material third-party dependencies behind important business services. The narrow point for outsourced security operations is in paragraph 6.5: where a firm uses third parties such as managed service providers or external security operations centres for monitoring, it should ensure they deliver equal or greater detection capabilities than the regulators expect of the firm itself. Logs should be immutable, kept secure, available to the regulators on request and retained for an appropriate period (paragraph 6.7), and paragraph 6.8 flags that AI agents with high autonomy may need granular logging of each step. Proactive threat hunting and deception or decoy techniques are expected where appropriate (paragraph 6.10).
Response and recovery: impact tolerance metrics beyond time
Chapter 7 starts from the operational resilience premise that disruption will occur. It expects a documented cyber incident response and recovery (CR&R) plan, which the draft defines as an umbrella term for every document relating to cyber response and recovery. No template is prescribed; a set of plans, policies, procedures and playbooks is acceptable if they are consistent (paragraphs 7.6 and 7.7). A firm in a group either keeps its own plan compatible with the group plan or ensures the group plan reflects its risks (paragraph 7.8). The plan should set out triage, activation and deactivation triggers, communications, measurable response and recovery objectives aligned to impact tolerances, the assets and capabilities needed, and the assumptions and accepted limitations behind it (paragraph 7.9).
Three external links are spelled out. Paragraph 7.15 asks firms to consider reporting a cyber incident to the NCSC “in addition to any applicable reporting obligations to the regulators and other authorities”, so an NCSC report is a separate decision from a SUP 15.18 or PRA notification. Paragraph 7.16 expects the plan to cover coordination with collective incident response frameworks, including the Sector Response Framework owned by the Cross-Market Operational Resilience Group (CMORG) and the Financial Services Cyber Collaboration Centre. Paragraph 7.17 asks firms to consider the links between their recovery and that of central banks, FMIs, critical third parties, AI providers and open-source communities.
The recovery section reopens impact tolerance design. Under the Operational Resilience Parts, as SS1/21 explains, every impact tolerance must include a time-based metric and firms must consider combining it with additional metrics such as volume or value (SS1/21 paragraphs 3.6 and 3.11). Paragraph 7.29 of the draft observes that destructive, persistent attacks may make recovery within a short, time-based tolerance challenging, and asks firms to consider additional non-time-based metrics such as value, volume, critical activity, end-users and types of payments. The draft links those metrics to setting the parameters of a degraded service and the priority actions, such as critical payments, to complete in that state while full recovery continues.
For high-severity incidents, paragraph 7.31 expects firms to be able to recover in a clean environment from trusted and immutable backups, to rebuild critical technology from a known trusted state, using bare metal recovery as the example, and to demonstrate their rebuild capabilities to the board and the regulators. Paragraph 7.32 suggests testing backups in isolated recovery environments before restoration.
The draft also sets a closing condition. Under paragraph 7.33, a firm should not consider recovery complete until it has assessed residual risks to its important business services, customers, counterparties, data and regulatory obligations and has acted to manage them. A ticket closed when the service comes back up would not meet that test.
Assurance: recovery testing, CBEST and STAR-FS
Chapter 8 asks for a cyber assurance programme that tests capabilities individually and in combination, with measurable indicators such as reductions in incident detection times and speed of recovery (paragraphs 8.3 to 8.5). The direction of travel is set in paragraph 8.11: testing should not focus exclusively or unduly on protective and detective capabilities, and where proportionate, response and recovery testing should move beyond tabletop exercises to areas such as how quickly the firm could complete a bare metal rebuild.
Where proportionate, paragraph 8.13 asks firms to consider threat intelligence-led assessments, such as red team exercises or ransomware simulations. The frequency, nature and scope of any assessments should reflect the firm’s cyber threat exposure, operations and risk profile. Paragraph 8.14 separates the two UK schemes: CBEST is regulator-led and applies only to certain firms, although all firms should consider its published thematic findings, while STAR-FS is voluntary and designed for a wider range of firms. Our comparison of STAR-FS and DORA threat-led penetration testing covers how the two schemes interact for groups subject to both regimes.
Paragraph 8.15 adds an AI-specific test case: a firm considering AI to support an important business service should consider red or purple teaming against model or system evasion, data poisoning, prompt injection and model extraction. Paragraph 8.17 points firms to lessons from SIMEX and the Cyber and Operational Resilience Stress Test, and to incident management playbook exercises run by critical third parties that supply them. Paragraph 8.12 asks that, as far as possible, testers are functionally independent of the teams that built and use the assets, with external testers where appropriate and proportionate.
Paragraph 8.19 ends the chapter by asking firms to evolve from point-in-time assurance towards automated assurance where possible. The draft does not say what level of automation supervisors would treat as sufficient, which is one of the areas Question 3 invites comment on.
The AI effective practices, and the AI text already inside the draft SS/FG
The AI document states at paragraph 1.2 that its practices are “not part of the regulators’ formal expectations”. The consultation paper places it outside the consultation (paragraphs 1.8 and 2.21), explains that this lets the regulators update it periodically as the technology changes, and invites comments. Paragraph 1.6 of the AI document says the practices are intended to be used proportionately.
Its content is operational. The practices that translate most directly into a control or contract change:
- A shift from periodic vulnerability fixing towards continuous exposure management (paragraphs 1.9 and 2.1).
- Risk-based remediation targets for priority vulnerabilities, with remediation speed and vulnerability backlogs treated as operational risks (paragraphs 3.4 and 3.5).
- A review of material third-party contracts for disclosure of embedded AI, service levels for vulnerability remediation, and prompt notification of changes to AI systems that may affect the third party’s cyber risk profile (paragraph 6.5).
- Compensating measures for open-source and open-weight tools, which give no direct recourse to a specific third party (paragraph 6.6).
- A review of identity and access management where the firm receives advanced third-party AI systems on a pre-release or restricted basis, including insider misuse (paragraph 7.3).
- Recovery steps specific to AI, such as rolling back fine-tuned models or agent states, redacting unsafe memory or embeddings, and retraining on uncompromised data (paragraph 8.5).
- Human oversight matched to risk: human-in-the-loop approval where errors could be serious or irreversible, human-on-the-loop for lower-risk activity, human-in-command arrangements to set autonomy and escalation, and kill switches that let systems be paused or shut down safely (paragraph 10.2).
- Forums such as the Frontier AI Information Sharing Forum and CMORG for coordinated responses to machine-speed attacks (paragraph 12.1).
The non-binding label covers this document only. The draft SS/FG, which is under consultation and would carry supervisory weight once final, already contains its own AI expectations: AI agents count as users for access control (paragraph 5.4), AI use cases belong in board MI (paragraph 3.16), unauthorised AI falls under shadow ICT (paragraph 3.21), AI inventories feed the asset inventory (paragraph 4.4), autonomous agents may need granular logging (paragraph 6.8), AI providers belong in recovery planning (paragraph 7.17) and a firm considering AI to support important business services should consider AI-specific testing (paragraph 8.15). A gap analysis that files every AI point under “non-binding” will miss those seven.
The regulators’ announcement cites a Bank of England and FCA survey in which 88% of respondents expected frontier AI cyber capabilities to have a moderate or high impact on their risk profile over the next two years, and 64% said vulnerability management and patching had seen the greatest increase in management attention. Annex 2 of the consultation also lists the FSB’s 2026 toolkit on responsible AI adoption, discussed in our FSB AI sound practices governance gap analysis.
Quantum effective practices and the NCSC migration milestones
The quantum document carries the same status statement at paragraph 1.2. Its starting point is the cryptographically relevant quantum computer (CRQC), a machine able to undermine public-key cryptography. None is known to exist, and paragraph 2.2 gives three reasons to start anyway: migration to PQC takes time, “harvest now, decrypt later” attacks collect encrypted data today for decryption later, and progress may be faster than expected, so firms may consider a range of plausible timelines when planning.
The governance and planning practices are concrete:
- Board and senior management agreement of a documented migration plan with milestones and dates, and resourcing for a multi-year programme (paragraph 3.1).
- Board MI on readiness, migration progress, inventory development, priority systems and third-party readiness (paragraph 3.5).
- Identification of cryptographic methods across applications, databases, APIs, certificates, public key infrastructure, authentication, backups, networks and operational technology, recorded in a cryptographic inventory that maps third-party dependencies (paragraphs 4.1 and 4.2).
- Prioritisation by data sensitivity and protection lifetime, system criticality, legacy constraints and implementation time (paragraph 4.4).
- Crypto-agility, the ability to change algorithms as standards evolve (paragraph 5.2), and hybrid schemes treated, in line with NCSC guidance, as an interim step towards a PQC-only end state (paragraph 5.5).
- Tests that failed changes can be reversed and that systems do not revert to vulnerable methods (paragraph 6.2), and recovery arrangements that cover compromised cryptography while keeping services within impact tolerances (paragraph 6.3).
Third parties follow the accountability line in paragraph 2.7 of the draft SS/FG, which ties it to the outsourcing and third-party risk framework. As non-binding effective practices, paragraphs 6.4 and 6.5 suggest that firms may seek information on material third parties’ migration plans and may need to review or update procurement processes and contractual arrangements relating to cryptographic dependencies and PQC readiness. Paragraph 6.6 states that firms remain accountable for managing quantum-related cyber risk where third parties provide the relevant systems or services.
The quantum document contains no migration deadline. For dates, paragraph 4.6 points firms to the latest NCSC guidance, which may change. The NCSC’s PQC migration timelines, first published on 20 March 2025, set three milestones: by 2028, define migration goals, carry out a full discovery exercise and build an initial plan; by 2031, carry out the early, highest-priority migration activities and refine the plan into a full roadmap; by 2035, complete migration of all systems, services and products. The NCSC aims that guidance primarily at large organisations, critical national infrastructure operators and companies with bespoke IT, while stating that the core timelines are relevant to all organisations. For an international comparison, our note on the HKMA PQC toolkit for banks covers Hong Kong’s approach.
The consulted text on quantum is shorter. Paragraph 5.11 of the draft SS/FG asks firms to follow established cryptographic standards, including PQC standards where relevant, and paragraph 4.4 names cryptographic bills of materials as an inventory source. Those two sentences are inside the consultation.
Cost estimates and the seven consultation questions
The PRA and FCA did not consult their cost benefit analysis panels, because no new rules are proposed and the annualised central cost estimate falls below the threshold (paragraph 4.17). Their central estimates:
| Population | One-off cost | Annual ongoing cost | Annualised over ten years |
|---|---|---|---|
| PRA-regulated firms, including dual-regulated (427 firms) | £14.9 million | £2.5 million | £4.2 million |
| Bank-regulated FMIs (10) | £730,000 | £58,000 | £140,000 |
| FCA solo-regulated firms (around 1,700) | £7.8 million | £1.7 million | £2.6 million |
Together these come to about £7.0 million annualised. Per firm, the FCA estimates one-off costs of £25,000 for a large firm, £7,900 for a medium firm and £1,400 for a small firm, with ongoing costs of £3,700, £1,700 and £500. The estimates are primarily staff-driven, covering familiarisation, gap analysis, training and process changes, and exclude existing capabilities. Paragraph 4.28 states that the regulators do not consider the expectations in themselves would entail costly new technology projects.
A PRA sensitivity case with 50% more staff effort raises its annualised figure to about £6.2 million. The benefits case is a break-even calculation: avoided losses of 3.5% of an illustrative £200 million annual cost of data-breach incidents (33 reported incidents in 2024 multiplied by an IBM average of £6.05 million) would offset the costs, which the paper presents as a threshold and does not forecast.
The consultation asks seven questions:
- Q1: general comments on the principles-based, outcomes-focused, technology-neutral approach.
- Q2: specific comments on Chapters 3 to 8.
- Q3: areas where more detail on expected outcomes would help.
- Q4: the approach to proportionality by size, business model, risk profile and systemic importance.
- Q5: the proposal to refer to the NCSC CAF.
- Q6: interoperability with other standards and frameworks.
- Q7: the cost benefit analysis, including the line between incremental costs and costs of existing obligations or firms’ own activity.
I read Q7 as the place to put on record any expectation a firm could only meet through a technology project, such as a clean-room rebuild capability, because the cost estimates assume the expectations themselves need none. The regulators will publish a general summary of responses and name respondents only with consent (paragraph 2.7 of the consultation paper’s section on responding).
Frequently Asked Questions
We are a UK subsidiary of an EU banking group already applying DORA. Does that satisfy the draft SS/FG?
It counts as supporting evidence, which is the most the draft offers. Paragraph 2.6 lets firms use compliance with non-UK standards as supporting evidence, but the firm must be able to show how it meets the UK expectations even where group policies are set outside the UK. Paragraph 3.4 adds that the UK board should keep enough expertise to manage the subsidiary’s cyber risk, check that group policies consider UK requirements including the SS/FG, and manage risks from intra-group ICT services.
We are an authorised payment institution supervised only by the FCA. Are we in scope?
Yes, if you are subject to SYSC 15A. The scope list in paragraph 1.9 names payment institutions, e-money institutions and registered account information service providers among the SYSC 15A firms covered. Annex 2(a) also lists regulation 98 of the Payment Services Regulations 2017 on management of operational and security risks among the requirements the guidance would be used to supervise.
If we report a cyber incident to the NCSC, have we notified the regulators?
The draft treats these as separate steps. Paragraph 7.15 frames NCSC reporting as something to consider in addition to any applicable reporting obligations to the regulators and other authorities, such as SUP 15.18 for FCA firms from 18 March 2027 or the PRA’s incident reporting rules under PS7/26. A firm may also need to give the NCSC indicators of compromise so it can support or monitor the incident.
Will we have to submit a CAF self-assessment or a new cyber return?
The consultation proposes no new rules and no new return. Firms are asked to use the CAF to develop and assess the maturity of their arrangements, the regulators would use it as a benchmark in supervisory engagement, and documentation is to be provided to the regulators on request (paragraph 3.13). Separate reporting obligations, such as operational incident reporting from 18 March 2027, come from other instruments.
Can comments on the AI and quantum documents change them before the final SS/FG?
The regulators welcome comments and say feedback on the two documents can be submitted with the consultation response. Because the documents are not formally consulted on and are designed to be updated periodically, the consultation paper gives no timetable for revising them.
Related Articles
- SS2/21 Outsourcing: The PRA Register and Notification Guide: the PRA’s outsourcing and third-party risk statement that the draft SS/FG relies on for material third-party cyber risk.
- STAR-FS and DORA TLPT: Threat-Led Testing for Firms in Both Regimes: how the voluntary UK STAR-FS scheme compares with DORA threat-led penetration testing.
- CPMI-IOSCO Cyber Resilience Toolkit: What FMIs Should Review: the toolkit the draft SS/FG cites for example cyber resilience metrics.
- HKMA PQC Toolkit: 23 Workstreams, Four Tools and a 2028 Checkpoint: Hong Kong’s post-quantum preparedness toolkit for banks, a comparison point for the UK quantum practices.
- UK Critical Third Parties Regime: 13 July 2026 Go-Live: the UK oversight regime for critical third parties referenced in the draft’s recovery and exercising expectations.
- SM&CR Phase 2: PRA CP14/26 Proposals, Responses Due 22 January 2027: the parallel PRA consultation on Senior Manager Functions, relevant to where cyber accountability sits.
Key Takeaways
- Response deadline: Friday 26 February 2027, to both CP17_26@bankofengland.co.uk and GC26-9@fca.org.uk; AI and quantum comments can go in the same response.
- Settle where cyber accountability sits: for SM&CR firms, SMF24 ordinarily, or another SMF such as SMF4 where that person has the authority, experience and functional independence the draft describes; firms outside SM&CR, including FMIs, allocate it under paragraph 3.20.
- Re-cut board cyber MI into current, planned and implemented-and-validated, and add AI use cases and third parties without cyber assurance.
- Align proposed board escalation criteria for major cyber incidents with both the operational incident reporting requirements under PS7/26 or SUP 15.18 and the broader notification expectations under PRA Fundamental Rule 7 or FCA Principle 11.
- For important business services exposed to destructive cyber scenarios, test whether a non-time-based metric (value, volume, critical activity) defines the degraded service.
- Check outsourced monitoring contracts against the paragraph 6.5 test of equal or greater detection capability.
- The quantum practices describe a cryptographic inventory that maps third-party dependencies (paragraphs 4.1 and 4.2); the NCSC’s first milestone, by 2028, is to define migration goals, carry out a full discovery exercise and build an initial migration plan.
Sources and References
- Bank of England / PRA, CP17/26 and GC26/9: Cyber risk and resilience management (19 November 2026): bankofengland.co.uk
- FCA, GC26/9: Cyber risk and resilience management, consultation page: fca.org.uk
- FCA GC26/9 / PRA CP17/26, joint consultation paper with the draft SS/FG (Annex 1) and the list of relevant regulation and guidance (Annex 2) (PDF): fca.org.uk
- PRA, FCA and Bank of England, Effective practices: Cyber risk and resilience management relating to artificial intelligence (AI) (19 November 2026): bankofengland.co.uk
- PRA, FCA and Bank of England, Effective practices: Cyber risk and resilience management relating to quantum computing (19 November 2026): bankofengland.co.uk
- FCA press release, UK financial watchdogs clarify cyber expectations as attacks rise (19 November 2026): fca.org.uk
- Bank of England news release, Cyber resilience guidance consultation and AI and quantum effective practices (19 November 2026): bankofengland.co.uk
- PRA, PS7/26 Operational resilience: Operational incident and third-party reporting (18 March 2026; implementation 18 March 2027): bankofengland.co.uk
- FCA, PS26/2 Operational incident and third party reporting: fca.org.uk
- PRA, SS1/21 Operational resilience: Impact tolerances for important business services (PDF): bankofengland.co.uk
- PRA, SS2/21 Outsourcing and third party risk management: bankofengland.co.uk
- FCA, Operational resilience (SYSC 15A firms): fca.org.uk
- NCSC, Cyber Assessment Framework: ncsc.gov.uk
- NCSC, Timelines for migration to post-quantum cryptography (20 March 2025): ncsc.gov.uk
- Bank of England, 2025 CBEST thematic (20 January 2026): bankofengland.co.uk
- Bank of England, Ensuring the resilience of CCPs: near-final policy statement (16 November 2026): bankofengland.co.uk
Gap analysis before the 26 February 2027 close
The draft SS/FG asks for no new submission, so the work it creates sits inside documents firms already maintain. Three of them carry most of the change: the board MI pack (paragraphs 3.14 to 3.16), the CR&R plan and the impact tolerance metrics behind it (paragraphs 7.9 and 7.29), and the asset inventory supported by material third-party register data under the reporting requirements taking effect on 18 March 2027 (paragraph 4.4). A chapter-by-chapter mapping of those documents to the CAF principles is the artefact that turns the consultation into a response, and it has to be finished before 26 February 2027 for any finding to reach the regulators through Questions 2, 3 and 7.
Disclaimer: The information on RegReportingDesk.com is for educational and informational purposes only. It does not constitute legal, regulatory, tax, or compliance advice. Always consult your compliance officer, legal counsel, or the relevant supervisory authority for guidance specific to your institution.
