FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Aipedia: Training providers and educators
AI proctoring can flag possible cheating in online exams, but Ofqual doubts that AI alone can reliably confirm work is the candidate's own. Wrongly voided results lead to appeals, discrimination complaints and claims.
Part of: AI in training providers and educators
In short
AI remote proctoring watches candidates through webcam, microphone and screen and flags behaviour that might suggest cheating. Ofqual notes the high bar needed to trust that remote assessment evidence is attributable to each learner, doubts that sole use of AI meets it, and flags bias and fairness concerns for different groups. Automatically voiding results is the main risk: it can disadvantage disabled candidates and others, and may engage data protection rules on automated decisions. Human review of every flag, reasonable adjustments and a clear appeal route are the core controls.
Online exams may be used by professional bodies, training providers and corporate academies. AI remote proctoring tools watch the candidate through a webcam and microphone, monitor the screen and flag behaviour such as looking away, another voice in the room, a second face or switching windows. Some tools verify identity using face matching. Flags may be reviewed by a person, or in some set-ups they lead automatically to a result being voided.
The attraction is cost and convenience. The problem is that an exam result can affect a candidate's job, licence to practise or immigration status, so a wrong flag has real consequences.
Ofqual's approach to regulating the use of AI (republished 16 July 2026) notes the high bar required to ensure that all remote assessment evidence can be trusted as attributable to each learner. It has said it doubts that sole use of AI meets that bar, and points to the need for appropriate human involvement and safeguards. It also flags the potential for bias and impacts on fairness for different groups of students. For regulated qualifications, that means AI proctoring should support, not replace, human invigilation decisions.
Unaccredited and corporate assessments have no equivalent rule, but the same logic applies when a result matters to the candidate. Personal data captured during proctoring, including video and any biometric matching, is also subject to data protection law; the DfE's generative AI guidance for schools and colleges stresses care with personal data in AI tools.
Proctoring tools may perform less well for some candidates: those with darker skin in poor lighting, candidates with disabilities that affect movement or eye contact, people sharing a home with family, or those with unreliable broadband. If flags fall more heavily on particular groups, discrimination arguments follow; see our note on the Equality Act 2010 disability provisions.
Illustrative scenario: a provider runs a certification exam with AI proctoring and voids results automatically when the tool flags suspicious behaviour. A candidate with a condition causing involuntary movements is flagged and loses their result, delaying a job offer that depended on the certificate. They complain that no person reviewed the decision and no adjustment was offered.
Where a result is voided automatically, with no meaningful human involvement, data protection rules on significant automated decisions may also apply; see AI in admissions and progression decisions.
A claim that a provider negligently ran an exam or wrongly voided a result causing financial loss may fall within a trainers' PI policy, subject to its terms. Discrimination and data protection allegations are treated differently across wordings, and a data breach involving exam recordings may fall to cyber cover; see GDPR and data breach cover within PI. Supplier contracts may limit what you can recover from the proctoring company.
Apex arranges PI for training providers and assessment bodies. See the training providers Aipedia hub, relying on AI detection tools and learner data in AI tools.
Ofqual has said it doubts that sole use of AI can meet the high bar needed to trust that remote assessment evidence is attributable to each learner. It points to the need for appropriate human involvement and safeguards. In practice, a person should review flags and make the decision.
A trained reviewer should look at the recording and context, consider any agreed adjustments, and give the candidate a chance to explain before any result is voided. Record the reasons for the decision and tell the candidate how to appeal.
Yes. Webcam and screen recordings identify the candidate, and face matching may involve biometric data. You need a lawful basis, clear information for candidates, sensible retention periods and secure handling, including where a supplier processes the recordings for you.
Candidates will usually look to the provider that set the exam. You may have a contractual claim against the supplier, but software terms may cap liability. Check your contract and make sure your own PI describes your assessment activities accurately.
Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.
Start a PI quote Request a callbackApex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.
