On 28 July 2015, the US Government Accountability Office published A Framework for Managing Fraud Risks in Federal Programs. On 2 May 2023, COSO and the Association of Certified Fraud Examiners (ACFE) released the second edition of their Fraud Risk Management Guide.
On 1 September 2025, the UK offense of failure to prevent fraud came into force.
The first two give a fraud detection and prevention program its structure. The third adds a UK legal test when someone acting for the firm commits a fraud meant to benefit it or its clients.
What is a fraud risk management framework?
A fraud risk management framework is how an organization decides which frauds could hurt it, who answers for each, which controls stop or catch them and how it proves those controls work. COSO's internal control framework covers fraud in Principle 8: an organization "considers the potential for fraud in assessing risks to the achievement of objectives." The COSO and ACFE executive summary gives fraud its own assessment for a core reason: "This fundamental difference is intent."
One framework covers two exposures. One is fraud against the firm by customers and outsiders, such as account takeover. The other is fraud by people acting for the firm and meant to benefit it or its clients, which the UK offense targets. For the first, controls sit in the payment flow, and each decision needs a reason its owner can check later. In complead, formerly Sanction Scanner, the fraud prevention product ends each check in approve, decline, review or step up, and gives every outcome a reason.
Which public frameworks can you build on?
Four public texts can shape a framework, each for a different group.
Table 1. Public fraud risk frameworks by jurisdiction (as of November 2026)
| Where | Framework or rule | Binds | Structure | Since |
|---|---|---|---|---|
| Any organization | COSO and ACFE Fraud Risk Management Guide | Voluntary | 5 principles consistent with COSO's 2013 internal control components | 2 May 2023 (first edition 2016) |
| US federal programs | GAO-15-593SP | Leading practices; under 31 U.S.C. 3357, agencies follow OMB guidelines requiring a fraud risk evaluation and risk based controls | 4 components: commit, assess, design and implement, evaluate and adapt | 28 July 2015 |
| Large organizations, UK nexus | Economic Crime and Corporate Transparency Act 2023 (ECCTA) s199 and Home Office guidance | Criminal offense; reasonable prevention procedures are a defense | 6 principles | 1 September 2025 |
| UK listed companies under the Code | UK Corporate Governance Code 2024, Provision 29 | Comply or explain | Board declaration on the effectiveness of material internal controls | 1 January 2026 |
The Home Office principles are top level commitment, risk assessment, proportionate risk based prevention procedures, due diligence, communication (including training), and monitoring and review. A large organization meets at least two of three tests in the financial year before the fraud: turnover above £36 million, a balance sheet above £18 million, more than 250 employees; a parent applies group tests. The Serious Fraud Office (SFO) is one of four prosecutors that can bring the charge.
How do you build a fraud risk management framework, step by step?
Build it in five stages, in this order: governance, fraud risk assessment, prevention and detection controls, investigation and response, then monitoring and improvement.
Table 2. Five stages of a fraud risk management framework
| Stage | Input | Output on record | Who decides |
|---|---|---|---|
| 1. Governance | Board expectations, offense scope | Policy, program owner, a nominated person per subsidiary, risk tolerance | Board and senior management |
| 2. Fraud risk assessment | Typologies, channels, past cases, audit findings | Register with inherent and residual ratings, fraud risk profile, decisions not to act, with the approver named | Risk owners, then the board |
| 3. Prevention and detection controls | The fraud risk profile | Fraud prevention plan, preventive and detective controls, due diligence on associated persons | First line, with second line challenge |
| 4. Investigation and response | Reports, alerts, whistleblowing | Case record, root cause, corrective action, self report decision | Investigations lead and legal |
| 5. Monitoring and improvement | Losses, detection results, control tests | Evaluation results, reported deficiencies, next review date | Internal audit and the board |
The order follows the five COSO and ACFE principles, which the guide draws as an ongoing process; GAO applies monitoring and feedback across all four of its components.
Who owns fraud risk, and what governance does a framework need?
The board owns it. In the Home Office's words, "Responsibility for the prevention and detection of fraud rests with those charged with the governance of the organisation." COSO and ACFE Principle 1 asks for a program that reflects the board's expectations, and GAO adds a dedicated entity to lead it. For groups, the Home Office's examples include a nominated person responsible for fraud prevention in each subsidiary.
Picture a payments group with three subsidiaries. The board approves the policy and the tolerance. The head of fraud runs the anti fraud program, and each subsidiary's nominated person signs its part of the assessment. Where fraud and AML report to different heads, the split between fraud and AML teams has to say who holds a mule account case.
The Home Office says compliance with the Code "may contribute" to the defense but "is not sufficient, on its own".
How do you run a fraud risk assessment?
Run it with named risk owners, in six documented steps. Skipping it is hard to defend: "it will rarely be considered reasonable not to have even conducted a risk assessment," the Home Office says.
- List who could commit fraud. Agents, contractors, staff in sensitive roles, and the schemes each could run.
- Apply the fraud triangle. Opportunity, motive and rationalization: thin oversight, financial stress or a sales target, a belief that nobody is harmed. The guidance also asks whether emerging technology such as AI opens new routes.
- Rate the inherent risk. Likelihood and impact before any added prevention measures, with a written reason for each rating.
- Map controls and rate the residual risk. Compare it with the board's tolerance, which GAO defines as "the acceptable level of variation in performance relative to the achievement of objectives."
- Write the fraud risk profile, GAO's name for the documented risks, their likelihood and impact, the tolerance and the priorities.
- Record every decision not to act, with "the name and position of the person who authorised that decision."
The ACFE's 2026 report, Occupational Fraud 2026: A Report to the Nations, studies 2,402 cases whose investigations ran between January 2024 and September 2025. Victims had formal fraud risk assessments in 51 percent of them. With one, the median loss was $80,000 against $150,000 without, and the fraud ran a median 10 months before detection against 15.
What goes into a fraud risk assessment template?
One illustrative register row shows the fields:
- Risk: a sales agent overstates investment returns to win clients, a fraud meant to benefit the firm.
- Inherent rating: likely, high impact; agents earn commission.
- Control and residual rating: scripted disclosures and call sampling; medium.
- Owner and decision: head of distribution; accepted until sampling reaches 10 percent of calls, signed by the chief risk officer.
- Next review: 1 June 2027.
Whether fraud sits inside the AML risk assessment or beside it, it needs rows of its own. The Home Office lets a firm extend an existing assessment to the frauds in scope, while COSO and the ACFE warn that simply adding fraud to one may miss deliberate acts.
Which prevention and detection controls belong in the framework?
In the ACFE's 2026 data, all 18 anti fraud controls were associated with lower median losses and 17 with faster detection. Four had the largest reductions in both: management review, proactive data monitoring and analysis, a code of conduct and surprise audits.
COSO and the ACFE sort controls into preventive, "designed to avoid a fraudulent event or transaction at the time of initial occurrence," and detective, designed to find one after the initial processing. The Home Office asks for a fraud prevention plan "proportionate to the risk identified in the risk assessment".
For customer fraud, a fraud score rates how likely each payment, login or application is to be fraud, and thresholds decide what an analyst sees.
How far does due diligence on third parties have to go?
To anyone performing services on the firm's behalf. In the Home Office's example, Bank C uses Bank D for clearing, and Bank D, with its senior management involved, commits a fraud that benefits Bank C's clients. Bank D is an associated person, and Bank C could be liable unless a court decides its procedures were reasonable.
The guidance asks for proportionate, risk based due diligence on everyone who acts for the organization. Its list of means includes screening tools. For a supplier acting on the firm's behalf, a supplier due diligence form records company details, owners, services and sanctions exposure.
What should happen when fraud is found?
Investigate, fix the failed control, update the assessment and decide whether to report. COSO and ACFE Principle 4 asks for a coordinated approach to investigation and corrective action, and GAO for a plan on how the program will respond.
Picture an internal audit that finds a procurement manager approving invoices from a company the manager part owns. The plan names who opens the case and keeps the evidence, and who decides on dismissal and recovery. From there it follows the AML or fraud investigation path of triage, evidence and decision, and that evidence is what a report to the police or a regulator rests on. The root cause, one person approving alone, becomes a second approver control.
Under the SFO's corporate cooperation guidance, firms that report promptly and cooperate fully will, "barring exceptional circumstances", be invited to negotiate a deferred prosecution agreement.
How do you monitor and improve the framework?
Monitor what the Home Office lists: detection of fraud and attempted fraud, investigations, and whether prevention measures work. COSO and ACFE Principle 5 asks for ongoing evaluations of whether each principle is present and functioning. GAO focuses them on outcomes.
Picture a yearly review of a 40 row register: 6 rows (15 percent) have no current owner and 3 controls were never tested. Both go to the board as dated deficiencies.
The Home Office leaves the review interval to the firm. Assessments typically run "annually or once every two years", it says, with an earlier review when external factors call for one. A court may find an assessment that was not reviewed recently enough "not fit for purpose". A statutory audit alone "cannot constitute sufficient defence".
Before a threshold changed at review goes live, complead's rules engine backtests it on real history, over 30, 60 or 90 days. Each version is kept with who changed it, when and why.
What changed in 2026
- 1 January. Provision 29 of the UK Corporate Governance Code 2024 began to apply.
- 9 March. The UK's Fraud Strategy 2026 to 2029 cited the Economic Crime Survey 2024: 1 in 4 UK businesses with more than one employee hit by fraud in twelve months, an estimated 6.04 million instances.
- 29 June. Section 250 of the Crime and Policing Act 2026 extended senior manager attribution to all crimes, and ECCTA sections 196 to 198 were omitted.
- 13 October. The SFO opened a direct route for a firm's lawyers to discuss a self report before filing it.
This week: name the owner of fraud risk and one person per subsidiary, record who accepted each untreated risk, and put the next review date in the board calendar.