QAA drafts AI principles for assessment up to doctoral level
The second draft says AI bans and detection software are not enough, and asks for points where students show their own learning.

The Quality Assurance Agency for Higher Education (QAA) published draft sector principles for assuring academic standards in the age of AI on 19 October 2026. QAA wants them to apply from further education through to doctoral level. They focus on assessment design rather than particular tools, around one question: does an assessment genuinely show what a student has learned? The text is a second draft and has not yet been agreed or endorsed by QAA or any other body.
01Bans and detectors are not enough
The draft says outright prohibition of AI and over-reliance on AI-detection tools are not effective primary safeguards of standards. It notes that detection tools are widely reported to carry significant rates of false positives and false negatives. Instead, it favours valid assessment design with meaningful security actions, backed by enforceable regulations. The draft says detection should gather evidence of learning, for example through questions or evidence of process, rather than rely on detection software.
02Every assessment needs a stated AI position
Whether AI is permitted, expected or restricted, the draft treats this as a deliberate design choice that is explicitly justified and built into assessment rubrics. Where AI is restricted, the rationale should make clear that the restriction does not extend to assistive technologies or reasonable adjustments. The draft also calls it unfair to set rules that providers have no practical means of checking. Where AI use is permitted, assessment should verify the student’s direction and critical evaluation of that use.
03Points where you show your own work
Some summative elements, not necessarily every assessment, should verify individual attainment under conditions where AI use can be controlled or verified. The draft lists forms such as presentation and questioning, or a structured defence of an artefact. It also says students are assessed on how they develop, justify and refine their work as well as on the final product. QAA says the final version will be published later this year, with shared definitions of key assessment terms.
04What this means for your thesis
These are draft principles for providers, not rules for students, and your own regulations still apply. Even so, they show the direction QAA is testing with the sector. Ask your supervisor or graduate school for the written position on AI tools for your thesis. Check that it covers any assistive tools you rely on. Keep dated drafts, notes and a short log of any AI use. They show how your argument developed and help you defend your choices in a viva, if you have one. If your rules ask you to declare AI use, the thesis structure guide shows where the declaration page usually sits in the front matter.
Thesis structure: every section in order
Every part of a thesis in order, what each chapter holds, an outline example, and the outline set up in Word.
Every fact comes from QAA’s news item of 19 October 2026 and its second-draft principles page. Both were reopened on 6 November 2026. Independent reviewers compared each statement with the source wording. They checked that the draft status and the doctoral scope match what QAA says, and that the draft does not mention theses or vivas. Three rounds of wording corrections were applied before a fresh independent check passed.
Sources
- Draft sector principles (generative artificial intelligence) QAA, 19 October 2026
- Draft principles for assuring academic standards in the age of AI QAA, 19 October 2026