Academic Integrity and Ethical AI Guidance for Colleagues

1. Purpose and Scope

This guidance sets out what academic integrity and ethical practice mean in relation to Generative AI (GenAI), for all academic colleagues involved in teaching, assessment, feedback, and supervision.

It sits alongside and should be read together with:

• The University's Guiding Principles for Generative AI - the six principles underpinning all use of GenAI across the University.
• The University's AI Policy - the legal, governance, and compliance requirements for GenAI use.
• Generative AI in Learning and Teaching: Practical Ideas and Resources - worked examples and use-case ideas for using GenAI well in your practice.

This document focuses specifically on the integrity and ethical dimensions: what's expected of you as a colleague and why when GenAI is part of your teaching, assessment, feedback, or supervision.

2. Why This Matters

GenAI can improve the quality of teaching and save colleagues' valuable time. But because it can now produce plausible teaching materials, plausible marking feedback, and plausible answers, the line between 'AI helped me do this well' and 'AI did this instead of me' isn't always obvious in the moment. Academic integrity in the context of GenAI isn't about avoiding the technology - it's about making sure your own judgement, expertise, and accountability stay at the centre of what you produce and decide, and that students can trust what you tell them about your working practices.

3. Core Expectations

3.1 Use GenAI purposefully and critically

Use GenAI where it genuinely adds value to your teaching, assessment design, or research, not by default. Whatever GenAI produces - a lesson plan, quiz questions, feedback comments, or a marking rubric - should be critically reviewed and, where necessary, revised before you use it. The exception is when you deliberately give students raw, unrevised GenAI output specifically so they can critically appraise it themselves. In that case, being transparent about what you've done and why makes it appropriate.

3.2 Be transparent, and expect the same of students

If you have used GenAI in developing your teaching materials, marking approach, or feedback, tell students clearly, and explain why and how it was used. Attribute AI-generated content or data properly, in the same way you'd expect students to declare their own use. Holding yourself to a lower standard of transparency than you expect from students undermines the standard itself.

3.3 You remain fully accountable

GenAI does not take responsibility for a mark, a piece of feedback, or a teaching decision - you do. Any final mark or substantive feedback must ultimately be your own, and you must be able to defend it. Relying solely on GenAI to assess student work, or to interact with students in place of your own engagement, is not permitted.

3.4 Protect the human-centred core of teaching

GenAI should support, not replace, the parts of teaching that depend on human judgement and relationships, including personal feedback, mentoring, and direct interaction with students. This isn't a sentimental point; it's why students study with others rather than work through a syllabus alone.

3.5 Apply the same evidence standard to students that you'd want applied to you

Any suspicion that a student has misused GenAI must be based on clear, verifiable evidence from the work itself, not on assumptions drawn from a student's engagement patterns, attendance, or how capable you judge them to be. This matters particularly for students from widening participation backgrounds, who are more vulnerable to wrongful suspicion based on this kind of indirect evidence.

3.6 Think before you use a tool that the University hasn't vetted

Use institutionally supported GenAI tools wherever possible. If you use a third-party tool the University hasn't reviewed, you are personally responsible for checking its terms and satisfying yourself that it doesn't expose you, your students, or the University to data protection, intellectual property, or copyright risk before you put any information, materials or data into it.

4. Where the Line Sits - What's Not Permitted

Some uses aren't a matter of judgement, they're not permitted at all, because they remove the human element that academic integrity depends on:

• Relying solely on GenAI to prepare teaching or learning materials, without your own critical review
• Relying solely on GenAI to mark or assess student work
• Relying on GenAI to interact with students in place of your own engagement - for example, generating personalised feedback or communications you present as your own, without review
• Creating deepfakes or other content prohibited under the AI Policy's Usage Guidelines (Section 5)

5. Illustrative Examples

Example 1: Appropriate

A colleague uses GenAI to generate a first draft of quiz questions for a formative assessment, then reviews and edits every question for accuracy and appropriateness before using them. Students are told the quiz was AI-assisted.

Why: a critical review happened (3.1), and it was disclosed (3.2).

Example 2: Not appropriate

A colleague uses GenAI to generate detailed feedback comments on a batch of essays and releases them to students without reading the essays themselves or checking the feedback.

Why: this crosses the 'interact with or assess students' line in Section 4; no real human judgement was applied to the individual student's work.

Example 3: Appropriate, with care

A module leader deliberately gives students a GenAI-generated case study containing factual errors and asks them to identify and correct them as a critical-thinking exercise, telling them upfront that the material is AI-generated and why.

Why: this is the deliberate 'raw output' exception in 3.1 - disclosed, and pedagogically purposeful.

Example 4: Not appropriate

A colleague pastes a batch of students' dissertation drafts into a free, non-institutional AI tool to get quick feedback suggestions before a supervision meeting.

Why: this puts personal student data into an unvetted third-party tool - a potential data protection and confidentiality breach, regardless of good intent (3.6).

Example 5: Not appropriate

A colleague notices a student's writing style has changed and, without checking the submitted work itself for evidence, refers them for suspected AI misuse on that basis alone.

Why: this doesn't meet the evidence standard in 3.5 - suspicion must come from the work itself, not inference from unrelated signals.

6. Related Documents and Where to Get Help

• Guiding Principles for Generative AI
• AI Policy
• Generative AI in Learning and Teaching: Practical Ideas and Resources
• Questions or uncertain cases - please email genai@westminster.ac.uk

7. Sector Context and Further Reading

This guidance was developed specifically for Westminster, but it doesn't sit in isolation. The expectations above are consistent with and informed by wider sector thinking on academic integrity and GenAI. The following are not required reading, but may be useful for colleagues who want the broader context:

• QAA Academic Integrity Charter and GenAI resources - the UK's designated quality body's core guidance on academic integrity, including specific advice on assessment in the GenAI era.
• Jisc guidance on AI and assessment - practical sector guidance on wording and approach for AI-related assessment policy.
• Advance HE - How can Generative AI be used in Learning and Teaching