Since 1 September 2025, a large organisation can be prosecuted in the UK when an employee, agent, subsidiary or anyone performing services on its behalf commits fraud intended to benefit it, unless it can prove it had reasonable fraud prevention procedures. The offence is created by section 199 of the Economic Crime and Corporate Transparency Act 2023 (ECCTA) and follows the model of the Bribery Act 2010 failure to prevent bribery offence. This guide sets out who is in scope, what counts as an associated person, the base fraud offences, the six principles in the Home Office guidance, and how to build and evidence reasonable procedures. It is written for compliance, legal, risk, fraud and security leaders, and it covers a point the official guidance only touches: how automated systems, AI agents and the non-human identities they use fit into a fraud risk assessment.
Key takeaways
- The offence is in force from 1 September 2025 and applies to bodies corporate and partnerships wherever they are incorporated, provided the fraud has a UK nexus.
- Size is tested on the group: a body is large if it meets two of three thresholds (more than 250 employees, more than £36 million turnover, more than £18 million balance sheet total) in the financial year before the fraud, counted across subsidiaries including those overseas.
- No knowledge is required: the prosecution does not need to show that senior managers knew of or ordered the fraud, only that an associated person committed a base fraud offence intending to benefit the organisation or its clients.
- The only defence is procedures: the organisation must prove, on the balance of probabilities, that it had reasonable prevention procedures, or that it was not reasonable to expect any. Following the guidance is not a safe harbour.
- Automation is part of the risk: the Home Office guidance asks whether emerging technologies such as AI create new fraud opportunities, so pricing engines, sales bots and agents acting on the organisation's behalf belong in the risk assessment and the controls.
Scope: who and what is covered
The offence has four building blocks: a relevant body that is a large organisation, an associated person, a base fraud offence, and an intention to benefit. All four must be present, together with a UK connection.
Large organisations
Section 201 defines a large organisation as a relevant body that meets two or more of these conditions in the financial year before the year of the fraud (the Act calls it "year P"):
| Condition | Threshold | How it is measured |
|---|---|---|
| Turnover | More than £36 million | Companies Act 2006 Part 15 meaning for UK companies; pro rata for a short financial year (s201(3), (5)) |
| Balance sheet total | More than £18 million | Aggregate of assets at the end of the financial year, or a corresponding measure (s201(5)) |
| Employees | More than 250 | Monthly average of persons employed under contracts of service across year P (s201(4)) |
The Home Office guidance (section 2.1.2) says the criteria apply to the whole organisation including subsidiary undertakings, and turnover is aggregated across subsidiaries, including those outside the UK. Parent undertakings are dealt with in section 202. The Secretary of State can change the thresholds by regulations (s201(6)), so check legislation.gov.uk each year rather than relying on a figure in a policy document.
A subsidiary that is not itself large can still be caught: under s199(2), where an employee commits fraud to benefit that subsidiary and its parent is a large relevant body. A parent can be liable when a subsidiary's employee commits fraud intending to benefit the parent (s199(8)). The guidance is clear that a parent is not responsible for a subsidiary's unrelated activities.
Associated persons
Section 199(7) lists employees, agents, subsidiary undertakings and anyone who otherwise performs services for or on behalf of the body. Whether someone performs services is judged on all the circumstances, not only the contract (s199(9)). The guidance (section 2.3) gives practical boundaries:
| Typically associated | Typically not associated |
|---|---|
| Employees, including temporary staff and those in overseas offices | Suppliers that only provide goods |
| Agents with authority to act or contract for the body, while acting for it | External lawyers, valuers, accountants or engineers who provide a service to the organisation rather than acting on its behalf |
| Subsidiary undertakings acting corporately | Franchisees by contract alone (unless they provide services for the franchisor) |
| Partners of a partnership; outsourced sales, collections or claims handlers acting for the body | Anyone committing fraud in a purely personal capacity |
Base fraud offences
Schedule 13 lists the offences that can trigger liability, plus aiding, abetting, counselling or procuring them (s199(6)):
- Fraud Act 2006: section 1 fraud (by false representation, failing to disclose information, or abuse of position), section 9 participating in a fraudulent business carried on by a sole trader, section 11 obtaining services dishonestly.
- Theft Act 1968: section 17 false accounting and section 19 false statements by company directors (sections 17 and 18 of the Theft Act (Northern Ireland) 1969).
- Companies Act 2006: section 993 fraudulent trading.
- Common law: cheating the public revenue, and in Scotland fraud, uttering and embezzlement.
The list can be extended by secondary legislation (s200). Tax evasion facilitation is covered separately by the Criminal Finances Act 2017, and bribery by the Bribery Act 2010, so many organisations will run one combined economic crime risk assessment.
Intent to benefit, the victim exception and the UK nexus
- Benefit: the associated person must intend to benefit the organisation, or a person to whom they provide services on its behalf, such as a client. The guidance (2.4) says the organisation need not actually benefit, benefit need not be the main motive, and non-financial advantages count.
- Victim exception: there is no liability under s199(1)(b) where the organisation was, or was intended to be, a victim of the fraud (s199(3)). An employee who steals from the employer is a matter for internal fraud controls, not this offence. See the Occupational and Internal Fraud Guide.
- UK nexus: an act forming part of the fraud took place in the UK, or the gain or loss occurred in the UK (guidance 2.5). An overseas organisation can be liable for fraud targeting UK victims; overseas fraud with no UK connection is out of scope.
Key obligations: the six principles
The statute does not prescribe procedures. The defence turns on whether the procedures were reasonable, and the Home Office guidance (chapter 3) frames that around six principles. Prosecutors and courts will read it.
| Principle | What the guidance expects | Evidence to keep |
|---|---|---|
| 1. Top level commitment (3.1) | Board and senior leaders own fraud prevention, set the tone, name a responsible senior person and fund the work, including technology such as third-party due diligence tools | Board minutes, named accountable executive, budget, policy statement |
| 2. Risk assessment (3.2) | Documented, dynamic assessment of how associated persons could commit fraud to benefit the organisation, using opportunity, motive and rationalisation (the fraud triangle); reviewed annually or every two years, sooner after a trigger (3.2.7) | Risk register by associated-person typology, likelihood and impact ratings, review dates |
| 3. Proportionate risk-based procedures (3.3) | Controls matched to the assessed risks, including the risk that controls are overridden | Control library mapped to risks, test results |
| 4. Due diligence (3.4) | Risk-based checks on people and firms acting for the organisation, contract clauses and termination rights, M&A diligence on targets' fraud controls | Screening records, contract templates, onboarding approvals |
| 5. Communication, including training (3.5) | Policies communicated, training weighted to the highest-risk roles, independent whistleblowing channels with protection from victimisation | Training completion and assessment results, speak-up case log |
| 6. Monitoring and review (3.6) | Detection of attempted fraud (including data analytics and the scope for AI), investigation of incidents, review of whether controls work | Detection rule inventory, investigation files, lessons-learnt reviews |
The guidance also notes that it may be reasonable to have no procedures only in rare cases, and that a risk assessment a court finds out of date may not be fit for purpose.
What a fraud risk assessment should cover
- Typologies of associated persons: list each group (sales staff, call centre agents, introducers, outsourced collectors, subsidiaries, contractors writing code or content) and the frauds each could commit for the organisation's benefit.
- Opportunity: weak oversight, single-person control of pricing, invoicing or reporting, manual overrides, and new technologies. The guidance (3.2.1) asks directly whether emerging technologies such as AI create new opportunities.
- Motive: sales targets, bonus structures, financial stress, pressure to meet market expectations (3.2.2).
- Rationalisation: "everyone does it", "no one is harmed", resentment (3.2.3).
- Information sources: data analytics, prior audit findings, sector bodies, regulator enforcement, prosecutions and deferred prosecution agreements (3.2.4), plus foreseeable emergencies (3.2.5).
Timeline and current status
| Date | Event | Source |
|---|---|---|
| 26 October 2023 | ECCTA receives Royal Assent; section 199 enacted but not yet in force for the offence | legislation.gov.uk |
| 6 November 2024 | Home Office publishes the statutory guidance to organisations on the offence | GOV.UK collection page |
| 1 September 2025 | Offence comes into force (S.I. 2025/349) | legislation.gov.uk, GOV.UK |
| 10 October 2025 | Guidance collection updated to use the term "subsidiary undertaking" | GOV.UK collection page |
As checked on 8 November 2026, section 199 is in force with no outstanding amendments shown on legislation.gov.uk. The offence applies only to frauds committed on or after 1 September 2025. Because thresholds and the list of base offences can change by regulations, re-check the statute and guidance at each annual review.
Enforcement and penalties
- Penalty: a fine (s199(12)). On conviction on indictment there is no statutory maximum; in Scotland and Northern Ireland a summary fine is capped at the statutory maximum. There is no custodial sentence because the defendant is the organisation, though individuals who commit the base fraud face prosecution in their own right.
- Prosecutors: the Crown Prosecution Service and the Serious Fraud Office in England, Wales and Northern Ireland (with the Public Prosecution Service in Northern Ireland), and the Crown Office and Procurator Fiscal Service in Scotland (guidance 2.7.1). Some regulators can also prosecute.
- Deferred prosecution agreements: available for this offence in England and Wales, not in Scotland or Northern Ireland, according to the guidance.
- Self-reporting: the SFO's corporate co-operation guidance applies; co-operation and disclosure are considered when deciding whether to prosecute or offer a DPA.
- Burden of proof: the prosecution must prove the base fraud by the associated person to the criminal standard; the organisation then carries the burden of proving reasonable procedures on the balance of probabilities.
See the Online Fraud Regulation Map for how this fits with other fraud duties.
What it means for AI agents and non-human identities
The Act talks about persons, and an AI system is not an associated person in its own right. The risk is that a human associated person uses, configures or instructs an automated system to commit fraud for the organisation's benefit, or that an outsourced provider's automation does so while performing services. The guidance's own questions about AI (3.2.1) and about using AI to detect fraud (3.6.1.1) make this a reasonable thing for a court to expect in the risk assessment.
| Automated scenario | How it maps to the offence | Reasonable control |
|---|---|---|
| Sales or support chatbot makes false claims about a product to close sales | Possible false representation by the employees who designed, prompted or approved it, or by the vendor operating it on the organisation's behalf | Approved claims library, output monitoring, human sign-off for changes to prompts and policies |
| Pricing or billing automation inflates charges or hides fees | Possible fraud by abuse of position or failing to disclose, by the staff who set the rules | Change control on pricing logic, segregation of duties between rule authors and approvers, reconciliation analytics |
| Agent with write access to finance or reporting systems alters figures | Possible false accounting by whoever directed or concealed it | Scoped, attributable machine identities, immutable audit logs, no shared service accounts |
| Outsourced provider's bots generate fake reviews or claims for the organisation | Provider performs services on the organisation's behalf, so may be an associated person | Due diligence on providers' automation, contract clauses, audit rights |
- Attribution: every agent, bot and integration should run under its own identity with an accountable human owner, so that an action can be traced to the person who configured or invoked it. Shared credentials destroy the evidence an investigation and a defence both need. See the Non-Human Identity (NHI) Guide and the Agent Identity and Access Guide.
- Least privilege and separation of duties: an agent that can both create and approve a payment, or both change and publish financial data, recreates the single-person control that the fraud triangle calls opportunity. See the Segregation of Duties Guide.
- Monitoring: log agent decisions and tool calls in a form that can be reviewed by fraud and compliance teams, not only by engineers, and include agent activity in the detection analytics required by principle 6.
Compliance steps
- Confirm scope. Test the group against the s201 thresholds for the last financial year, including overseas subsidiaries, and record the result. Identify subsidiaries that are large in their own right.
- Appoint ownership. Name an accountable senior executive, agree a board reporting line and budget, and decide whether to fold the work into an existing economic crime programme (principle 1).
- Map associated persons. Build an inventory of employee groups, agents, intermediaries, outsourced service providers and subsidiaries, and of automated systems that act on the organisation's behalf with the humans accountable for them.
- Run the risk assessment. For each typology, record plausible frauds that would benefit the organisation or its clients, rate inherent risk by likelihood and impact, and identify existing controls (principle 2).
- Close gaps proportionately. Prioritise high inherent risks: incentive design, approval thresholds, segregation of duties, claims and marketing review, change control on pricing and reporting logic (principle 3).
- Update due diligence and contracts. Add fraud prevention clauses, audit and termination rights to agent and outsourcing agreements; screen high-risk roles and partners (principle 4). See the Third-Party Risk Management (TPRM) Guide.
- Train and communicate. Deliver targeted training to high-risk roles, refresh the speak-up policy and channels (principle 5). See the Cyber Whistleblowing Guide.
- Monitor, test and review. Deploy analytics on high-risk processes, test controls, investigate allegations with a documented process, and review the assessment at least every two years and after any incident or significant change (principle 6).
Small vs large in-scope organisations: a group just over the thresholds can rely on a concise risk assessment, a short policy and a handful of targeted controls; a multinational with intermediaries in many markets needs local risk owners, a nominated fraud prevention person in each subsidiary (as the guidance suggests in 2.6) and analytics across business lines. Organisations below the thresholds are not caught directly, but large customers will push the requirements down through contracts.
Measuring whether procedures work
- Percentage of associated-person typologies with a current, dated risk assessment (target 100%, none older than 24 months).
- High-risk role training completion and assessment pass rate.
- Third parties acting on the organisation's behalf with completed due diligence and fraud clauses in contract.
- Number of speak-up reports on fraud for the organisation's benefit, and median time to close investigations.
- Automated systems acting externally with a named accountable owner and reviewed audit logs.
Framework mapping
| Six principles | NIST CSF 2.0 | NIST SP 800-53 Rev. 5 |
|---|---|---|
| Top level commitment | GV.RR-01, GV.RM-01 | PM family (programme management) |
| Risk assessment | GV.OC-03, ID.RA-03 | RA-3 |
| Proportionate procedures | GV.RM | AC-5 Separation of Duties |
| Due diligence | GV.SC-06 | PS-3 Personnel Screening, SR-6 Supplier Assessments and Reviews |
| Communication and training | PR.AT-01 | AT-2 Literacy Training and Awareness |
| Monitoring and review | DE.CM-03 | AU-6 Audit Record Review, Analysis, and Reporting |
Common failure patterns
- Copying the bribery programme: a renamed anti-bribery policy misses the frauds that matter here, which sit in sales, marketing, reporting and claims rather than in gifts and hospitality.
- Testing size on one entity: measuring the UK company alone and ignoring overseas subsidiaries' turnover, which the guidance says must be aggregated.
- Ignoring the "benefit to clients" limb: wealth managers, brokers and agencies focus on fraud against themselves and miss fraud their staff commit for clients.
- Incentives left untouched: sales targets and bonuses that reward outcomes regardless of how they were reached are the motive in the fraud triangle; an assessment that does not review them is weak evidence.
- Automation with no owner: bots and agents running under shared service accounts with no log review make it impossible to show who configured a misleading process, or to demonstrate monitoring.
- A stale risk assessment: a one-off exercise in 2025 that is not reviewed after reorganisations, acquisitions or new products, which the guidance warns may be found unfit.
Practitioner checklist
- Record the s201 threshold test for the group, including overseas subsidiaries, each financial year.
- Name the accountable senior executive and report fraud prevention to the board at least annually.
- Maintain an inventory of associated-person typologies and of automated systems acting on the organisation's behalf, each with an owner.
- Complete a documented fraud risk assessment using opportunity, motive and rationalisation, and review it at least every two years.
- Review incentive schemes, approval thresholds and segregation of duties in high-risk processes.
- Add fraud prevention, audit and termination clauses to agent, intermediary and outsourcing contracts.
- Give each bot, agent and integration a unique identity, least privilege and reviewed audit logs.
- Deliver targeted training to high-risk roles and test that speak-up channels work.
- Keep an evidence pack mapped to the six principles, ready for a prosecutor or auditor.
Standards and references
- legislation.gov.uk: Economic Crime and Corporate Transparency Act 2023, section 199
- legislation.gov.uk: Economic Crime and Corporate Transparency Act 2023, section 201 (large organisations)
- legislation.gov.uk: Economic Crime and Corporate Transparency Act 2023, Schedule 13 (fraud offences)
- Home Office: Offence of failure to prevent fraud introduced by ECCTA
- Home Office: Guidance to organisations on the offence of failure to prevent fraud
- NIST: Cybersecurity Framework 2.0
- NIST: SP 800-53 Rev. 5
Related NHI Mgmt Group resources: Fraud Prevention Maturity Model · Occupational and Internal Fraud Guide · Insider Threat Programme Guide · Online Fraud Regulation Map. This guide summarises the UK failure to prevent fraud offence and is not legal advice.