Policy & Regulation

Ofqual to Redesign A Level Assessments Over AI Authorship Risks

Ofqual will propose redesigning non-exam assessment in A Level history and English, saying AI makes authorship of unsupervised extended writing impossible to reliably verify.

Ofqual Plans A Level Assessment Changes Over AI Misuse Risks
Ofqual Plans A Level Assessment Changes Over AI Misuse Risksernestoeslava / pixabay
By Sophie Lindqvist5 min read

Why it matters

  • Ofqual letter published November 18, 2026 proposes changing non-exam assessment in A Level history and three A Level English qualifications.
  • Invigilator and centre staff training on wearable-technology misuse must be in place for the summer 2027 exam series; responses due by November 30, 2026.
  • Almost half of roughly 5,000 annual malpractice cases over three years involved phones or other communication devices in exams.
  • 2025 penalties included 1,125 students losing an entire GCSE or A Level and nearly 2,000 cases of deducted marks.
  • Chief Regulator Sir Ian Bauckham: "Where we can no longer have sufficient confidence in the authenticity of assessed work, we must consider whether the assessment itself needs to change."

Ofqual will propose changes to non-exam assessment in A Level history and three A Level English qualifications because artificial intelligence has made it "increasingly difficult to establish reliably" whether unsupervised work is the student's own. Chief Regulator Sir Ian Bauckham set out the plan in a letter to leaders of awarding organisations published on November 18, 2026. The intervention applies to England.

The move signals something bigger than a crackdown on cheating. Bauckham's position connects malpractice controls directly to assessment design: when the conditions around an assessment no longer provide reliable evidence of authorship, detection alone cannot fix the problem.

What did the regulator actually say?

Bauckham's concern centers on work produced without supervision, particularly extended writing. He said in the full letter that Ofqual can no longer be confident about the authenticity of such work in some subjects.

"Where we can no longer have sufficient confidence in the authenticity of assessed work, we must consider whether the assessment itself needs to change," Bauckham wrote.

That quote carries the whole argument. The regulator isn't saying every non-exam assessment is broken — Bauckham explicitly acknowledged that not all non-exam assessment is inherently vulnerable to AI. He identified particular risks in some forms of assessment, and he named the ones that worry him most: A Level history and the three A Level English qualifications.

Ofqual will also consider changes in other subjects and qualifications where risks to authenticity require action. The regulator has invited awarding organisations to share their views on its direction before any formal consultation begins. In plain terms: the final rules aren't written yet, and the boards get a say first.

Why does this matter beyond exam halls?

Non-exam assessment — coursework, extended projects, essays written outside invigilated conditions — has long been a pillar of A Level grading in humanities subjects. It measures skills that timed exams capture poorly: sustained argument, independent research, extended writing itself.

If Ofqual redesigns or removes those components in history and English, the shape of how hundreds of thousands of students are graded in core humanities subjects changes with it. The stakes run in two directions at once. Keep unsupervised essays, and the qualification risks resting on work whose authorship nobody can verify. Remove them, and the exam system loses one of its few structured measures of independent, long-form thought.

The open question Bauckham leaves on the table is which forms of extended writing can still reliably measure a student's own achievement when the work happens without supervision.

What about wearable technology?

The November 18 letter doesn't stop at coursework. It also demands stronger action against misuse of wearable devices in exams, which are becoming more sophisticated and AI-enabled.

Awarding organisations must take "all reasonable steps" to protect assessment integrity and ensure required assessment conditions are implemented effectively and consistently. Relevant centre staff, including invigilators where appropriate, are expected to receive effective training built on clear, consistent and practical resources to identify and respond to wearable-technology misuse.

The deadline is concrete: that training must be in place for the summer 2027 exam series. Bauckham also made clear the burden doesn't fall on invigilators alone — organisations were asked to consider additional preventive and detection measures beyond what happens in the exam room.

Responses are due before November 30, 2026. Bauckham asked each organisation to describe:

  • Its risk assessments
  • Preventive action already taken or planned
  • Support provided to centres
  • Any further measures it considers necessary

Organisations planning collective action were also asked to explain how they will coordinate effectively with other awarding organisations.

What do the malpractice numbers show?

This intervention has a paper trail. Bauckham's earlier letter of March 3, 2026 requested stronger measures against mobile phones and other devices in exam halls, and against AI misuse in non-exam assessment.

That March letter carried hard figures on penalties in 2025:

  • 1,125 cases in which students lost an entire GCSE or A Level qualification
  • Nearly 2,000 cases in which marks were deducted
  • Roughly 5,000 malpractice cases reported annually over the preceding three years

Almost half of those roughly 5,000 annual cases involved students taking mobile phones or other communication devices into exams. The physical device, not the essay generator, was already the system's biggest enforcement problem before AI entered the picture.

Detected AI misuse cases reported to Ofqual remained relatively low at the time of the March letter. But the letter recorded significant concern among teachers, leaders and others about the real extent of misuse — a gap between what gets caught and what actually happens that regulators increasingly treat as the central problem.

How did we get here?

Before the March letter, Ofqual had already asked exam boards to improve awareness of what constitutes misuse and to strengthen deterrence, detection and prevention. That earlier guidance included more rigorous candidate and teacher authentication of work.

The November letter tightens the screws in two specific ways. It names the A Level subjects — history and the three English qualifications — where assessment design itself may change. And it sets a specific training timetable for wearable risks: practical invigilator preparation by summer 2027.

The trajectory is worth noting. Ofqual moved from general deterrence advice, to device and AI enforcement demands in March, to redesigning assessments themselves in November. Each step concedes a little more ground to the technology.

What happens next?

Awarding organisations have until November 30, 2026 to respond, after which Ofqual's formal proposals will follow — covering non-exam assessment in the named subjects, with other qualifications potentially drawn in where authenticity risks demand it.

The direction Ofqual has chosen extends beyond policing individual misconduct. The real test of the forthcoming proposals is whether any form of unsupervised extended writing can survive as trusted evidence of what a student can do on their own.

Source: unite.ai

Share this article:

More from Sophie Lindqvist

Sophie Lindqvist

Show full bio

Staff writer covering regulation, policy and their impact on artificial intelligence and machine learning at AI In Context.

749 articles

Related articles

  1. England's OfS Tells Universities to Grade Students Without AI
  2. Harvey Completes PROTECTED-Level IRAP Assessment for Australian Agencies
  3. England's Top Civil Judge Says AI Should Decide Simple Court Cases
  4. A Former Professor Estimates 90% of His Online Students Used AI for Coursework
  5. Bank of England governor demands 'right to intervene' as AI risk grows

« Previous article