Plagiarism and AI at Cambridge: acknowledged use is allowed

The short answer

Cambridge allows you to use AI in assessed work as long as you acknowledge it — unless your assessment brief says otherwise. That is a more permissive default than most universities, and the exact opposite of Oxford’s.

It is also the best-documented institution in this series: it publishes its sanctions, its standard of proof, and how many cases it actually brings each year.

The rule, in one sentence

From the Education Quality and Policy Office, this is the whole operative rule:

“A student using any unacknowledged content generated by artificial intelligence within a summative assessment as though it is their own work constitutes academic misconduct, unless explicitly stated otherwise in the assessment brief.”

Two hinges. The offence is unacknowledged use — so acknowledged use is not the offence. And the assessment brief can override in either direction, which makes the brief the controlling document.

For anything that is not summative, Cambridge is explicitly open:

“Students are permitted to make appropriate use of artificial intelligence tools to support their personal study, research and formative work. Where doing so, it is recommended that you discuss this with your supervisor or lecturer to understand how best to engage with these tools whilst still benefiting from the educational experience as intended.”

Cambridge publishes no prescribed acknowledgement wording, so make it clear and specific — which tool, for what, on which part. Our AI disclosure statement guide covers the shape.

The opposite of Oxford

The two ancient universities have landed on opposite defaults, and if you have read about one it is easy to assume the other works the same way. It does not.

• Cambridge: acknowledged use is permitted unless the brief forbids it. The failure mode is silence about your tools.
• Oxford: use is prohibited unless specifically authorised in advance, and unacknowledged use is separately plagiarism. The failure mode is either one.

We cover the Oxford regime in its own page. If you have friends at the other place, do not take their rules as yours.

There is no blanket screen

Cambridge uses Turnitin UK, but deliberately does not run everything through it:

“Although not mandatory, the University encourages all Faculties and Departments to make use of Turnitin software, although the method of use will vary depending on the subject area and course; some courses will screen all work, while some may undertake random checks or only screen individual pieces of work if concerns are raised.”

The examiners’ guidance says the same from the other side: “it is not a University requirement for examiners to use this software routinely. It is also possible that individual examiners or assessors may identify possible copied text or other forms of academic misconduct independently of detection software.”

Two things follow. Whether your work is screened depends on your course. And a concern can arise from a person reading your work, with no software involved at all — which is worth remembering if you have been assuming the only risk is a percentage.

Students do not get access as of right: “Students are not provided with access to the licence, except where this is provided by the course or College as part of a formative exercise on their course of study.” If your course offers a formative Turnitin exercise, take it.

You can remove your work afterwards

This is unusual and worth knowing. Cambridge does not ask for consent to check your work — but it gives you a right afterwards:

“The University does not require consent from students to submit work to the Turnitin database for the purposes of academic misconduct detection and education […] However, you do retain the right to remove your work from the Turnitin database after it has been checked, if you wish. We hope that you will choose to keep your work within the database so as to maximise the effectiveness of the software and to protect your work from future attempts to plagiarise it.”

Note the argument Cambridge makes for staying in — your work in the database also protects you from someone else copying it later. That is a fair point and worth weighing before you opt out.

There is also a privacy protection built into the terms: “Material submitted directly to Turnitin UK will use only examination numbers, or another number created for that purpose; student names will not be used, including in the filename.” Your work goes in under a number, not your name.

One honest limit: the policy page links to an IT support page explaining how to exercise the removal right, and that page did not resolve when we looked. So we can tell you the right exists but not the mechanism — ask your Faculty or OSCCA for the current route.

What a detection report can and cannot do

Cambridge publishes no statement about whether Turnitin’s AI-writing indicator is switched on in its tenancy — the Turnitin policy page describes similarity matching only, and we will not guess.

What it does publish is how such a report may be used, and this is the useful part. The examiners’ guidance:

“Reports generated by text-matching or detection software require careful interpretation, preferably by a subject expert. A report without interpretation is not proof of academic misconduct; it simply highlights text that appears to match with other text already contained within its database.”

And its evidence checklist for a suspected AI case expects more than a report: the concern form, your submitted work, “detection software reports from AI detection tools or other reasons to suspect AI”, plus “analysis from assessor/examiner regarding the likelihood of inappropriate use of AI”, plus your own-work statement and the guidance you had been given.

So a score is one item among six, and must be accompanied by a human analysis. The student-facing page makes the same point about false matches: “The matched text can often include a number of entirely innocent matches, such as entries in the bibliography, the essay title used by all students, or small matches like ‘the University of Cambridge’.”

That is the right way to read any such number, ours included: our published error rates are measured on English and German corpora, and no detector produces proof. See what a Turnitin score actually means and how accurate AI detectors really are.

A documentary note, since we are being precise about sources: the examiners’ guidance file is version 1.2 dated 18 December 2025, but its internal running footer still reads v1.1, September 2023. The scope clause covers assessments submitted from 1 October 2023 onwards.

The numbers Cambridge publishes

Cambridge is one of only two institutions in this entire series with usable, current case data. From the OSCCA Annual Report covering 1 October 2024 to 30 September 2025:

• 78 academic misconduct cases investigated (69 the year before).
• 67 found to be a breach (55 the year before); 11 no breach found.
• 69 decided by a Department or Faculty alone, 6 appealed, and 3 referred to the Discipline Committee.
• 6 appeals received, 3 upheld. Split by level: 51 undergraduate, 27 postgraduate.

Read them as counts for one year at one university, not as a rate — they cannot tell you your own odds of anything. What they do show is shape: the overwhelming majority of cases are settled at Faculty level with educational or mark-based outcomes, and only three reached the body that can touch a degree.

The sanction mix bears that out. At Chair-of-examining-body level in 2024-25: 42 cases got “a mark reflecting the unaffected parts”, 20 an educative or reflective session, 11 a mark of 0 for the affected assessment, 10 a written reflection, 4 a capped re-sit. At Discipline Committee level, exactly one case resulted in “amendment of academic results or temporary or permanent removal of academic awards”.

Three tiers of sanction

The Student Disciplinary Framework, in force from 1 October 2023, sets three levels.

Chair of Examiners, Senior Examiner or Chair of Degree Committee may impose: an educative session on academic integrity; a mark reflecting only the unaffected parts; a mark of 0 for the affected assessment; an apology; a written reflection; or, where re-sits are permitted, a capped re-sit.

The Student Discipline Officer may additionally require payment of material damages up to £250, a written apology, an educative or reflective session, a written reflection, or a no-contact order.

The Discipline Committee may impose restrictions on access to University or College premises and services; “the amendment of academic results or the temporary or permanent removal of academic awards”; temporary or permanent exclusion from membership of the University; and restrictions on representing the University.

There is a proportionality constraint published alongside it: the Committee “shall impose the lowest sanctions commensurate with the breach”, and “sanctions affecting the academic results or academic awards of a Respondent shall only be imposed where it is proportionate to do so.”

If you are on a research course

Cambridge publishes worked examples, and the two that concern a thesis are worth reading side by side because the line between them is where the stakes change.

Small amounts— “using small amounts of unauthorised or unacknowledged sources or materials within their thesis or dissertation […] limited to small amounts, or found within certain sections, such as the introduction or sections which do not focus on a student’s ability to analyse, criticise or otherwise complete the objective of the assessment (this can include self-plagiarism)” → resubmission with a capped mark, or marking that discounts the affected text, plus an apology, a written reflection and an educative session.

Significant amounts, or the analysis itself — “using unauthorised sources or materials to produce significant amounts of content within a thesis or dissertation […] or producing sections of analysis or core arguments within the assessment” → referral to the Discipline Committee.

The distinction is not really about volume; it is about which part. Material in an introduction is treated differently from material in your analysis, because the analysis is the thing being assessed. Note too that self-plagiarism appears in both examples.

A second offence also goes straight to the Discipline Committee. And the guidance flags its own limit: it applies to taught students, while suspected misconduct by postgraduate research students “while rare is very serious”, and anything touching an award already granted “is likely to require referral to a Discipline Committee”.

The evidence has to prove it

Cambridge’s standard of proof is published, and the second half of it is the part worth carrying:

“The standard of proof used when making determinations under this procedure is on the balance of probabilities. […] Decisions must be supported by evidence; it is not enough to simply believe that something is likely to have happened. This requirement means that there may be some cases in which the University decides that it is not appropriate to take or continue action under this procedure.”

And the examiners’ guidance puts the burden explicitly:

“It is for the evidence to prove that such misconduct has taken place, rather than the student to prove that it did not happen.”

You are not required to prove you wrote it. That said, bringing your drafts and dated version history is still the fastest way to end a conversation — see how to prove you wrote it yourself and what to do when you are accused.

If a detector score is what started this, our guide to an AI detector flagging your thesis sets out what that number does and does not establish on its own.

You have 10 working days to appeal a Faculty decision to the Appeal Committee, a panel of three trained persons that can “amend, quash or impose sanctions”. Once internal procedures are exhausted, a Completion of Procedures letter opens the route to the Office of the Independent Adjudicator. Unlike Oxford, Cambridge sets no deadline for reporting misconduct in the first place.

Before you submit

• Read the assessment brief. It can override the default in either direction and is the controlling document.
• Acknowledge any AI use in summative work. That single act is what separates permitted use from misconduct.
• Talk to your supervisor about formative use. Cambridge recommends it and it costs nothing.
• Take the formative Turnitin exercise if your course offers one — it is the only student access there is.
• Decide about the database opt-out consciously, knowing that staying in also protects your work from being copied.
• Be especially careful in your analysis chapters. That is where the published tariff escalates.

If you want a read on your own text before you hand it in, our AI check is free up to 1,500 words, with no account and no name.

Sources

• University of Cambridge: Artificial intelligence — Education Quality and Policy Office; the operative AI rule.
• University of Cambridge: Turnitin — screening practice, the anonymity requirement and the removal right; terms agreed by the General Board’s Education Committee, June 2021.
• University of Cambridge: Student Disciplinary Framework and Procedure — in force from 1 October 2023; sanctions and standard of proof.
• University of Cambridge: OSCCA Annual Report 2024-25 — the case data quoted above.

This page is orientation, not legal advice. What binds you is your assessment brief and the University’s disciplinary framework; your College runs its own separate discipline.