Why ACCA withdrew remote exams and why the retreat is not a fix.

Institutions abandon remote assessment fairly often. Most of them are small, were improvising, and never had serious controls in place. None of that applies here, which is what makes the decision worth reading closely.

So this was not a body without options or without scrutiny. It had the category-leading remote invigilation platform deployed, two regulators supervising it, and the failure mode already named on its own risk register. It also had a direct financial disincentive: examinations are more than a third of its income, and centre-based delivery constrains volume in a way remote delivery does not.

ACCA publishes its Disciplinary Committee decisions, and two from 2025 describe the failure mode precisely. Both concern live, remotely invigilated exams. We have not reproduced the candidates' names: they are identifiable individuals, ACCA publishes those decisions as a regulator exercising its own functions, and the evidential value here lies entirely in the mechanism rather than in who was involved. The linked decisions carry the full detail.

The webcam defines a frame. Everything the candidate needs sits outside it. A phone below the desk, a second laptop to the side, a tablet on a stand behind the monitor: none of these are exotic, and none of them require the candidate to defeat the proctoring software at all. They simply require the proctoring software to be looking at one thing while the candidate uses another.

The rules had not caught up either

There is a second, quieter gap. ACCA's Exam Regulations contain no clause naming artificial intelligence, generative AI, or LLM tools. AI use has to be prosecuted under the generic "unauthorised materials" and "electronic devices" headings. ACCA's published list of student misconduct types (cheating, collusion, contract cheating, impersonation, misrepresentation, plagiarism) does not include AI use as a category at all.

This is not a criticism unique to ACCA; most exam regulations in circulation were written before the tools existed. It does mean that an institution relying on rules alone is enforcing against a category its own rulebook does not define.

Moving candidates back into buildings does close the specific gap that remote proctoring could not. It is also expensive, exclusionary, and, for an exam body, structurally awkward in a way that shows up directly in its own accounts.

It costs more

It excludes candidates

Reported ACCA student impact includes 280-mile round trips within the UK, with centre gaps in Northern Ireland, Devon, Cornwall and Surrey. Every mile of that is a candidate who might not sit, and the effect is not evenly distributed: it falls hardest on candidates who are rural, working, caring for someone, or without a car.

And volume is the variable the finances rest on

ACCA's own going-concern disclosure sets out what happens if exam volume falls far enough. The reverse stress test, built to identify the conditions under which ACCA would exhaust its financial reserves, required a 60% reduction in exam volumes; in that scenario ACCA depleted its cash reserves in October 2026.

Sixty per cent of exam income is £53.7m. In the same year, operating surplus fell 93%, from £13,933,000 to £914,000, as strategic investment rose 59% to £40.2m.

This is why the retreat reads as a stopgap rather than a settlement. It was adopted because the alternative on offer, trusting an unmanaged device with a camera pointed at it, had stopped being credible. It is not obviously stable, and nobody involved has claimed it is permanent.

Every figure on this page comes from an audited financial statement, a regulator's published enforcement notice, an official disciplinary decision, or the institution's own website. Nothing is estimated. Where an inference is ours rather than a published finding, it is marked in the body of the report.

What we deliberately excluded

• The €200,000 Italian proctoring GDPR fine. The Garante fined Bocconi University €200,000 in 2021 over proctoring software, but the Court of Milan substantially accepted the university's appeal and reduced the fine to €10,000, holding that the image collection was not biometric processing. Quoting the original figure without that history would be misleading, so we have left it out.
• The KPMG Canada $700,000 PCAOB fine (March 2025). That action concerns inaccurate Form AP filings, not exam cheating. It appears alongside the genuine cases in search results and is excluded from the $197.9m total.
• Any remote-versus-centre volume split for ACCA. ACCA has never published one, so the exact number of sessions relocated by this decision is unknown and we have not modelled it.
• Any count of AI-cheating cases behind the decision. No source, ACCA's own included, has published one. The decision itself is the evidence, not a case count.
• ACCA's aggregate disciplinary totals. The 752 complaints and 148 concluded Disciplinary Committee cases reported for 2024 span all conduct types, including CPD defaults, ethics and PER breaches. ACCA does not publish exam malpractice as a separate category, so we cite the individual decisions instead.

Primary sources

Spotted an error? Our editorial policy sets out how we handle corrections, and hello@aiseptor.com reaches the author directly.