AI Use in Assessment
- Version
- 1.3
- Policy owner
- Lead IQA / Compliance Manager
- Approved
- November 2026
- Next review
- April 2027
The policy
DAIS-POL-023
AI Use in Assessment Policy
Data and AI School of London, NCFE Approved Centre
Policy Owner: Lead IQA / Compliance Manager
Version: 1.2 | Date: 20 October 2026 | Review: April 2027 (or upon any NCFE/JCQ or OTHM AI guidance update)
NCFE Risk Level: High
Mode of Delivery: Fully Online
23.1 Purpose and Scope
The Data and AI School of London occupies a distinctive position: it teaches Artificial Intelligence while operating in an environment where AI tools can be misused to undermine assessment integrity. This policy draws a clear and enforceable line between the legitimate use of AI as a curriculum tool and its misuse in assessments for NCFE and OTHM qualifications.
This policy complies with JCQ and NCFE guidance on AI in assessment (from September 2025) and is aligned with the school's Malpractice and Plagiarism Policy (Policy 10).
For OTHM qualifications, this policy also follows OTHM's Managing Use of Artificial Intelligence in Assessment Centre Guidance (version 2.0, September 2026). Section 23.8 sets out the rules for OTHM qualifications. Where section 23.8 differs from sections 23.2 to 23.7, section 23.8 applies to OTHM qualifications and sections 23.2 to 23.7 continue to apply to NCFE qualifications unchanged.
23.2 The Core Distinction
AI as a learning tool: Learners are expected to understand, build, and use AI systems as part of their qualification. Use of AI tools in learning activities, laboratory exercises, and exploratory tasks is encouraged. AI as an assessment tool: Submission of AI-generated text, code, data analysis, reports, or other content as original assessed work, without explicit permission for that specific task, constitutes malpractice under Policy 10.
23.3 Task-Level AI Permission Framework
For every formally assessed task, the assessor specifies one of the following permission levels in the assessment brief, published on the VLE:
| Level | Label | Meaning |
|---|---|---|
| 0 | No AI | No AI tools may be used at any stage. The work must be entirely the learner's own. |
| 1 | AI for research only | AI tools may be used to search for information or generate ideas. All submitted work must be written by the learner. |
| 2 | AI-assisted | AI tools may assist with drafting or coding. Use must be declared. Final work must be substantially the learner's own. |
| 3 | AI-integrated | The task specifically requires AI tools as part of the deliverable. Full use is expected and must be documented. |
Submitting work at a higher level of AI use than the stated permission level constitutes malpractice and will be investigated under Policy 10. For OTHM qualifications, see 23.8.3.
23.4 Declaration Requirement
For all assessed submissions at Level 0, 1, or 2, learners must include a completed AI Use Declaration confirming: whether AI tools were used and which; how they were used; and that the final submission is substantially their own work and that they understand and can explain all parts of it. Submissions without a completed declaration will be held pending clarification from the learner. For OTHM qualifications, see 23.8.4.
23.5 Detection and Investigation
Assessors are trained to identify indicators of AI-generated content, including:
Uncharacteristically polished or formal language inconsistent with the learner's previous writing style
Generic, non-contextualised content that does not reference the learner's own experience
Factual inaccuracies or "hallucinations" characteristic of large language model outputs
Code that is structurally correct but contains variable names or comments that do not reflect the learner's own work
Where AI misuse is suspected, the matter is referred to the Compliance Manager under Policy 10. The learner may be asked to undertake a viva voce via video call to confirm their understanding of the submitted work.
23.6 Consequences
For NCFE qualifications, a grade of Not Yet Competent (NYC) for the submitted unit
A requirement to resubmit under supervised conditions
Disqualification from the qualification
For NCFE qualifications, referral to NCFE under JCQ Suspected Malpractice procedures
For OTHM qualifications, the outcomes set out at 23.8.7
23.7 Policy Maintenance
This policy is reviewed immediately following any update to NCFE, JCQ or OTHM guidance on AI in assessment. Where guidance changes materially, all currently enrolled learners are notified via the VLE and required to re-confirm their understanding before their next assessed submission.
23.8 OTHM Qualifications
This section applies to every OTHM qualification the school delivers. OTHM units are graded Pass or Refer. Refer is the outcome recorded on OTHM's marking sheets, and the specifications describe the same structure as Pass / Fail.
23.8.1 AI and AI misuse
AI includes standalone and embedded tools that generate, rewrite, translate, summarise, analyse or otherwise transform content, including AI functions built into word processors, search engines, learning platforms and other software.
Every assessment decision must be supported by valid, sufficient and authentic evidence of the individual learner's achievement. AI must not perform the knowledge, analysis, evaluation, judgement, calculation, communication or practical competence that the learner is required to demonstrate. Acknowledging AI use does not by itself make that use acceptable.
Inappropriate or unauthorised use of AI is learner malpractice, in the same way as plagiarism, collusion and contract cheating, and it is dealt with under Policy 10.
23.8.2 Acceptable and unacceptable use
Whether a use of AI is acceptable depends on what the assessment is intended to measure, read alongside the qualification specification and the assessment requirements. Subject to those requirements, acceptable use may include:
Generating initial ideas where idea generation is not being assessed
Producing search terms for independent research
Creating a preliminary structure that the learner then develops independently
Limited spelling, grammar or structural corrections, where permitted, that do not substantially rewrite the work or misrepresent the learner's English language proficiency
Limited translation support, where permitted and where English language proficiency is not being assessed, provided the learner understands the submitted content
Obtaining an explanation of a concept as part of learning before assessment
Using an approved accessibility function or reasonable adjustment
Generating practice questions that do not form part of the submitted evidence
Permitted use must still be acknowledged where it has materially influenced the submitted work. Summarising source material, commenting on a draft, rewriting text for clarity, generating code, calculations, images or diagrams, analysing data, recommending evidence or conclusions, producing reflective prompts and translating substantial parts of assessed work are acceptable only where they do not replace the learner's assessed performance. If a tool has performed an activity covered by the learning outcomes or assessment criteria, the resulting content cannot normally be treated as evidence of the learner's achievement.
Unacceptable use includes:
Submitting AI-generated or AI-rewritten content as the learner's own
Using AI to produce the analysis, evaluation, judgement or reflection being assessed
Paraphrasing AI output to disguise its origin
Using AI to complete calculations, code, designs or practical decisions that the learner must perform
Fabricating, or failing to verify, citations, evidence, data or workplace examples
Failing to acknowledge material AI use, or giving an incomplete or misleading account of how AI was used
Using AI during an assessment where access is prohibited, or after being expressly instructed that AI is prohibited
Using AI to impersonate a learner in a discussion, presentation or practical assessment
Signing an authenticity declaration for work that is not the learner's own
AI must not be used to misrepresent a learner's English language proficiency, for example by changing limited or unclear English into fluent academic or professional writing.
23.8.3 Permission levels on OTHM tasks
OTHM's assignment briefs are OTHM's materials and the school does not alter them. For an OTHM task, the permission level under 23.3 is stated in the task instructions published with the brief on the VLE. A permission level is an assessment-specific instruction. It may narrow what 23.8.2 allows, but it never permits a use that 23.8.2 treats as unacceptable, and it does not add assessment criteria or change OTHM's requirements.
Every unit of the OTHM regulated qualifications the school delivers is set at Level 2, AI-assisted, decided by the Head of Centre on 4 November 2026 and stated in the task instructions on each unit's Assignment Brief page. AI tools may assist the learner's work, every use must be declared before submission, and the work submitted must be the learner's own. Level 3 is not used on those units: it describes a task that requires AI tools as part of the deliverable, which no unit of an OTHM regulated qualification does.
This does not apply to the school's endorsed learning programmes, which OTHM endorses and quality assures rather than awards. Those programmes teach the use of these tools as their subject, so their tasks do require AI tools as part of the deliverable and every one of their units is at Level 3, AI-integrated, as their own briefs state. DAIS-SD-045 sets out how the school describes endorsed and regulated provision, and the distinction matters here: 23.8 governs the regulated qualifications.
This level is read subject to 23.8.1 and 23.8.2, which prevail over it. AI must not perform the knowledge, analysis, evaluation, judgement, calculation, communication or practical competence the learner is required to demonstrate, and declaring a use does not make it acceptable. A learner declares each unit separately at the AI use declaration form, as 23.8.4 requires.
23.8.4 Formal submission and the declaration
Work is formally submitted when the learner declares it final and submits it for summative assessment.
Learners on OTHM qualifications give the AI Use Declaration with every formally submitted piece of work, at every permission level, and confirm that the work is their own. Where AI has materially contributed, the learner states which tool was used, what it was used for, the prompts or instructions given, the output produced and how the learner used, checked or changed that output. The learner keeps the prompts and a non-editable copy or screenshot of the output, and submits them where relevant. A bibliography entry naming an AI tool is not sufficient on its own.
At induction, and again before assessment, learners are told which uses of AI are permitted and which are prohibited, how AI use must be acknowledged, that evidence of permitted use must be kept, and the consequences of a false declaration or of misuse.
23.8.5 Authenticating work
Assessors become familiar with each learner's normal level of subject knowledge, use of terminology, written and spoken communication and approach to analysis and problem-solving, and follow the development of the work, for example through drafts, version histories, progress discussions and comparison of the final submission with earlier work. A baseline writing exercise may be used but is not treated as definitive.
Possible indicators of AI use include an unexplained change in style, quality or vocabulary; generic content that does not address the given context; technical language used incorrectly; false, unverifiable or irrelevant references; confident but inaccurate claims; inconsistency between sections; repeated formulaic structures; examples that do not match the learner's stated experience; work the learner cannot explain; metadata or version history inconsistent with the stated process; and remnants of prompts, chatbot instructions or generated disclaimers. None of these proves misuse. Each is a reason to make further enquiries.
Detection software is one part of the school's authentication arrangements. A probability or percentage is not proof that a learner used AI, and a detector result is never the sole basis for a finding of malpractice. Where a detector report is relied on as supporting evidence, the school keeps the tool and version used, the date of the check, the file checked, the complete report, the relevant passages, the human review undertaken and the other evidence considered.
An authenticity discussion, including the viva voce described at 23.5, uses open, non-accusatory questions specific to the submitted work, is recorded accurately, and verifies the existing evidence. It is not used as a reassessment or as an opportunity to supply missing evidence after submission.
23.8.6 Concerns before and after formal submission
A concern that arises before formal submission is dealt with through teaching and assessment: the learner is reminded of the rules, any further support stays within permitted limits, and a record is kept. It is not reported to OTHM unless it also involves suspected centre or staff malpractice, a wider security concern or another notifiable event. Feedback before submission must not allow repeated submission of AI-generated drafts until they evade detection.
Where misuse is suspected after formal submission, the submitted work is preserved in its original form. It is not replaced, amended, withdrawn or returned for informal correction while the concern is considered. The concern and the evidence that prompted it are recorded, the learner is told of the concern in neutral language and given a fair opportunity to respond, and the matter is investigated under Policy 10. Evidence that supports and evidence that contradicts the concern are both considered. A finding of malpractice is made only where the evidence shows, on the balance of probabilities, that malpractice was more likely than not to have occurred.
The investigation record and supporting evidence are retained under DAIS-POL-006 and are made available to OTHM promptly on request, including during external quality assurance. Suspected centre or staff malpractice is reported to OTHM promptly, at qualityassurance@othm.org.uk. Suspected or confirmed learner malpractice involving AI, identified after formal submission, is also reported to OTHM promptly, as DAIS-POL-010 section 10.8 sets out.
23.8.7 Outcomes
The NCFE outcome of Not Yet Competent is not used for OTHM qualifications. Work that cannot be authenticated is not accepted as evidence of achievement, whether or not intentional malpractice is established. The assessment decision is not confirmed, and certification is not claimed, until the investigation is complete and the school is satisfied that the evidence is valid, sufficient and authentic. Where malpractice is found, the outcomes in 23.6 that may also apply are a requirement to resubmit under supervised conditions, where reassessment is permitted, and disqualification from the qualification. The learner may appeal under the Appeals Policy (DAIS-POL-001).
23.8.8 Staff use of AI and data protection
Assessors and internal quality assurers remain responsible for their decisions. AI is not used as the sole or primary marker, to make the final decision on whether criteria have been met, to replace assessor or IQA judgement, to decide whether malpractice occurred, or to generate feedback that the responsible assessor has not reviewed and accepted. Where AI assists marking or feedback, the assessor reviews the complete submission, the school records how AI was used, and OTHM is informed during external quality assurance if AI has materially supported marking or IQA.
Learner work, personal information, confidential materials and OTHM intellectual property are not uploaded to an AI system unless that use has been properly authorised and complies with data protection, confidentiality and intellectual property requirements.
23.8.9 Training and review
Assessors and internal quality assurers are trained in the capabilities and limitations of common AI tools and of any detection software the school uses, the school's authentication procedures, how to investigate concerns fairly, and when a concern must be reported to OTHM. Learners are trained in the ethical use of AI, the risks of misuse and how to acknowledge AI use. The school reviews these controls at least annually and after any substantiated incident.
This policy is reviewed annually. Next review: April 2027 (or upon any NCFE/JCQ or OTHM AI guidance update). Approved by: Head of Centre. © Data and AI School of London, April 2026. NCFE Approved Centre.