Undergraduate Teaching 2026-27

Use of Generative AI Policy

Use of Generative AI Policy

Not logged in. More information may be available... Login via Raven / direct.

PDF versionPDF version

Approved by Faculty Board June 2026

Introduction

One of the aims of a university education is to encourage students to think to develop their cognitive abilities, and with excessive use of generative AI (Gen AI) there is a risk of outsourcing thinking and so not developing cognitive abilities. Notwithstanding this, it is appreciated that Gen AI can have benefits to student learning so this policy aims to address these competing concerns at a point when the Gen AI landscape is constantly changing.

This document sets out guidance for the use of Generative AI by students within the Engineering and Manufacturing Engineering Triposes. As it contains a mixture of rules and advice this document uses the word 'must' to identify rules where a failure to adhere might constitute academic misconduct and the word 'should' to identify best practice advice.

The landscape around Generative AI is continually changing, consequently this document sets out principles rather than attempting to provide fine grained advice covering specific situations.

General Principles

The requirements and guidance set out in this document have been based on the following principles.

  1. Students are responsible for everything they submit for assessment, irrespective of use of Gen AI. This includes every factual claim, all referencing and citations, all representation of data, all use of language and all arguments.
  2. Students will be expected to be able to explain and account for all aspects of submitted work, if a student is unable to do so this will be taken to imply that relevant learning outcomes have not been achieved. This will be reflected in any marks awarded even if there is no evidence of academic misconduct.
  3. The reflections of students on the use of generative AI are valuable, and these will be used to inform policy in the future. Students are therefore asked to specify the way in which they have used AI in submissions (unless otherwise stated). Such statements may also be used when assessing a student's original contribution to a piece of work.
  4. It is recognised that the use of generative AI has become very widespread. The Department is interested to see how it is being used, and does not wish to prohibit or penalise reasonable use.
  5. It is generally better to ask an AI tool for suggestions, which you can evaluate, rather than to ask the tool to 'do the work for you'.
  6. AI tools can be very helpful tools to assist in inclusion in education for students with disabilities or for whom English is a second language. The University provides access to NotebookLM which is a tool specifically designed to help organise complex information.
  7. The Department also wishes to minimise the chances of students inadvertently committing academic misconduct.

Requirements

  • In submitting work students must note the University's statement that '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.' More guidance is provided later in this document on academic misconduct issues.
  • Unless otherwise stated in an assessment brief students must declare all usage of Generative AI used in the preparation of a piece of summative work.
  • Individual assessment briefs will normally include specific AI guidance which may prohibit certain usage that would be normally acceptable, for example a foreign language exercise may prohibit the use of AI translation and grammar checking tools. Such per assessment guidance must be adhered to and supersedes any general rules or advice in this document.
  • Students must not use Gen AI to generate raw text and content for submission or for assessed presentations without working from an initial self-written base. The ability to use critical thinking to solve complex problems is a key competence for an Engineer.
  • Students must not use Gen AI to automatically generate and format references and bibliographies. Experience shows that this results in poor outcomes and it is an important requirement that where a reference is used the reference in question has been read to ensure relevance.' It is permitted to use tools to track and format references that the student has themselves identified, such as EndNote, Bibtex or similar.

Advice

The University Blended Learning Service offers the following general guiding principles for student use:

  • The use of Gen AI tools is not an effective replacement for developing skills or understanding in your subject area.
  • Gen AI tools can commonly produce incorrect or unsubstantiated information and as such should always be verified from trusted and reliable sources.
  • Consider thoughtful use of GenAI and associated software where possible, making appropriate use to support your own development, using the most effective tool for the task at hand, and using efficient prompt-engineering to reduce the amount of iteration necessary.
  • Remain aware of the limitations, inconsistencies, and biases that can exist within GenAI tools and data sets, and exercise caution when deciding to use information provided by software.
  • Be accountable and take responsibility for how, when, and why you decide to use generated materials or information from Gen AI software.

Additionally, the Department notes that :

  • It is unwise to ask a generative AI to produce anything (including ideas) that you do not feel competent to evaluate. There is no point asking for an output if you can't tell whether it has actually improved your work, understanding, et cetera.
  • The University makes available some AI tools via subscriptions (primarily Google Gemini and NotebookLM) and students may wish to consider using those models in preference to paid models. This reduces inequality in access and ensures that agreements on use of training data reached by the University apply.
  • Consider carefully if you are actually saving any time with a particular use of Gen AI, sometimes repeated cycles of 'prompt engineering' can end up taking more time than simply completing a task.
  • It's good to consider what you would be doing if you were not consulting an AI, and ask whether anything is lost by using the AI instead (e.g., the environmental harms of AI and the social benefits of talking to a well-informed friend).
  • 'AI can be a useful assistant, but not a reliable authority'
  • Generative AI can be an effective tool for researching background material and understanding concepts. Students should, however, bear in mind that, in a similar way to many technical Wikipedia articles, it can surface information that is unreliable yet sounds authoritative. Lecture notes and textbooks should be used to check such information.
  • Students should be aware that they could produce a piece of work with extensive permitted use of declared AI that, nonetheless, has not allowed the demonstration of key learning outcomes. Such work, even if correct, will not attract good marks. 
  • This document applies primarily to the use of Gen AI in summative exercises (those that are marked for credit). The use of Gen AI to complete formative work, primarily examples sheets, is not recommended as the process of solving problems (and indeed initially failing to solve problems) is an important way in which threshold concepts in learning are identified and addressed.' In the same way that copying cribs to submit to a supervisor is pointless and harmful, generating example sheet answers with extensive AI use means that little learning actually takes place.

Academic Misconduct

The University of Cambridge has issued the following statement with regard to the use of generative AI:

"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."

However, unacknowledged use of generative AI is not the only way that a student may commit academic misconduct. The University defines academic misconduct as:

"Academic misconduct is gaining or attempting to gain, or helping others to gain or attempt to gain, an unfair academic advantage in formal University assessment, or any activity likely to undermine the integrity essential to scholarship and research"

Taking these definitions together, then, a student might commit academic misconduct through use of generative AI by (even if unintentionally):

  • submitting anything produced by a generative AI for assessment without acknowledgement;
  • including, even unknowingly, fabricated or falsified data within the submission as if it were real;
  • including hallucinated sources and citations within the submission;
  • failing to correctly identify the source of a factual claim, idea or quotation (which may be because the generative AI did not correctly source the claim or idea).

Acknowledgements

This document was prepared based on the Department of Engineering's 'MLMI Gen AI Policy V1.0' and the University's AI policy Framework. Gen AI was not used in the preparation of this document.

 

Appendix: Example use cases in an Engineering Context

This appendix provides advice on a number of use cases for AI in an Engineering context. In all cases where a particular assessment provides assignment specific guidance that takes precedence over the contents of this Appendix. Red highlighting is used to identify use of Gen AI which may be inconsistent with the policy set out above, yellow highlighting is used to highlight uses where particular care is required.

 

Activity

Description

Example

Risks

Best Practice

Permitted /

Prohibited

1. Personal Reflection

Encourage structured reflection of learning, teaching, or research.

Seek reflective prompts, questions, or guidance and use them to inform personal development.

Use of GenAI may limit personal insight and might lose depth or become formulaic.

Maintain privacy when considering personal reflections and discuss regularly without AI assistance.

Permitted but talking to other students, supervisors and academics will likely lead to better understanding.

2. Time Management and Planning

Support the planning and organisation of assignments, deadlines and priorities.

Generate schedules, break down complex tasks, and prioritise tasks based on deadlines.

Schedules may not accurately reflect real-work constraints or suit what works best for you.

Treat suggestions as guides and apply your own workflow whilst regularly reassessing to ensure accuracy.

Permitted

3. Topic Selection

Assist in identifying potential research or project topics.

Suggest relevant areas of focus based on brief descriptions or assignment briefs.

Topics may be too broad, irrelevant, or not aligned with your specific goals or learning outcomes.

Use suggested topics as inspiration and refine based on personal interests and task parameters. Refer to own literature reviews and expert input to better identify your approach.

Permitted as a way of developing ideas, but students remain responsible for defending all ideas and concepts in their work and should be able to generate their own ideas

4. Understanding Key Concepts

Explain key concepts in more appropriate language and further expand relationships between ideas.

Reinterpret lecture notes or reading materials to supplement learning and wider discussion with peers.

Generated explanations may over-simplify or be inaccurate. You may risk reducing engagement and more critical understanding of key materials.

Always verify explanations against trusted sources and follow up with reflection or discussion with colleagues.

Permitted but discussions with staff, supervisors and other students will likely lead to better understanding.

5. Literature and Information Location

Identify relevant sources or datasets based on a provided area of interest or topic.

Suggest searching strategies, link to useful online resources and help to streamline research processes.

Suggested sources may be incomplete, inaccurate, and of variable quality. You may overlook important material if the search is ineffective.

Use AI as a guide, not a substitute for thorough research. Verify the quality and credibility of the suggested sources.

Permitted, but this should be viewed as a supplement for more traditional means of identifying sources. Review papers are an excellent and human curated source of references.

6. Identifying Trends in Literature

Identify gaps in research, support further strategic planning, and help to further contextualise research.

Analyse collections of literature or research to highlight recurring themes and emerging areas of interest.

AI may overemphasise superficial trends or overlook more subtle patterns. There is high risk of misinterpretation.

Critically evaluate outputs against original sources and treat identified trends as indicative and not definitive.

Permitted, but cannot substitute for forming own understanding of literature from own reading.

7. Idea Generation

Provide suggestions, prompts, or examples to stimulate ideas for projects and research.

Propose research methodologies, areas of investigation, or new lines of inquiry.

There is a risk of reducing originality and critical engagement, and for AI to provide generic and unhelpful suggestions.

Treat the suggestions as inspiration, not finished outputs. Critique and combine outputs with personal ideas and those of peers and experts.

Permitted to provide suggestions of 'where to go next' but not for generating text for submission.

8. Idea Interrogation

Evaluate, challenge and refine ideas to develop stronger arguments and develop reasoning.

Generate questions that test your current understanding, highlight assumptions and suggest alternative perspectives.

Questions and feedback may be superficial or misleading, this could reduce personal skills in critical thinking and evaluation.

Consider the questions as prompts for further thinking, use them in discussions with peers, and verify broader perspectives with additional research.

Permitted as an adjunct to discussion with other students, supervisors, staff etc.

9. Expanding Critical Thinking and Problem Solving

Support the analysis of complex problems and encourage the exploration of differing perspectives.

Suggest alternative approaches, common misunderstandings or assumptions, and challenge your reasoning.

Suggestions may oversimplify complex reasoning and introduce bias into your research.

Evaluate assumptions and reasoning critically, validate additional sources, and combine outputs with discussion and peers review to test ideas.

Permitted as an adjunct to discussion with supervisors, staff etc.

10. Generating Supplementary Materials

Assist in planning, designing and creating materials.

Generate resources, handouts, activities, research questions to support broader outcomes.

Outputs may not fully align with your intended outcomes, disciplinary standards or institutional policies.

Treat outputs as inspiration and not final design, check alignment with relevant standards and discuss with peers and consult with experts to validate.

Assignment dependent

11. Generating Examples or Case Studies

Provide illustrative examples or scenarios to test ideas, support research, and communicate findings.

Generate text or visual examples that help to communicate your work in various contexts.

Examples may be inaccurate, generic, or fail to capture the nuance of your intended output.

Verify examples against authoritative sources, adapt the examples to reflect specific contexts and clearly communicate where you have generated examples or identified real-world cases.

Assignment dependent. Should not be used to generate starting point for exercises where you are required to conduct own research to generate a case study.

12. Outline Creation

Support the structuring of projects, reports, and essays and begin with a clear foundation.

Generate writing frameworks, contents outlines, or suggested sequences.

There is a high risk of formulaic or generic structures and the limitation of your creativity in producing these outputs.

Vary approaches and use outlines as a starting point and adapt to meet your individual requirements, ensure a logical and relevant flow to the suggested structure.

Permitted for structuring the student's own ideas. The student must still be able to understand and defend the way that the work has been structured.

13. Generating Text

Produce drafts of sections, paragraphs, or illustrative content for essays, reports, or presentations.

Summarising texts to draft introductions or conclusions, writing whole sections of an essay or populating slides.

Generated content is susceptible to inaccuracies, lack of nuance, irrelevance and reflecting known biases. Over-reliance can reduce critical thinking and personal authorship.

 

Not permitted.

14. Paraphrasing

Reword existing text whilst maintaining original meaning and aid in clarity.

Rewrite your own drafts more concisely or use to reinterpret source material to support engagement.

AI may inadvertently alter the meaning of the text, oversimplify content, or produce text that raises ethical concerns about plagiarism.

Use sparingly to retain keen focus on materials, interrogate paraphrased text and always verify accuracy and meaning is maintained.

Permitted. Students should be aware of the danger of losing their own individual 'voice'. Must not be used to enable plagiarism of the work of others.

16. Translating the Writing of Others

Translating content by others for understanding, review, or analysis.

Broadening research to consider perspectives from other languages and cultures.

Accuracy and context of original sources may be compromised and could lead to ethical concerns.

Verify translations or use trusted sources and maintain appropriate attribution and ethical use of translated content.

Permitted in most circumstances but may be forbidden in exercises that directly assess language skills

17. Summarising Notes and Reading Materials

Condense longs texts into concise summaries to engage with wider materials.

Summarising own lecture notes, highlight key points, provide concise abstracts of reading lists.

Summaries may omit important information or misrepresent ideas. Over reliance can reduce engagement with the original source materials.

Use summaries as a starting point, not a replacement for engaging with the text yourself. Reflect on identified highlights and patterns to deepen understanding.

Permitted

18. Data Analysis

Assist with organising, summarising and identifying patterns in qualitative and quantitative data.

Provide insights, highlight trends, suggest visualisations, and provide additional sources to validate or challenge findings.

Analyses may misinterpret data, overlook contexts, or produce biased results.

Always cross-check AI outputs with standard analytical methods and ensure the context and parameters or analysis are understood.

Permitted.

Caution should be taken not to input personal data of others (e.g. from a survey) into an LLM without permission and consideration of data protection implications.

19. Mathematical Problems

Support understanding of complex processes, mathematical concepts, and check understanding.

Provide worked examples, solutions, or step-by-step guidance for mathematical tasks.

AI solutions may be incorrect, incomplete, or mislead through improper explanation or reasoning. AI often describes common methods despite not using them in its approach.

Verify solutions independently and with expert input, understand the reasoning behind AI generated results and use AI to support your own working, not replace it.

Assignment dependent.
Not recommended for formative supervision work.

20. Code Generation

Support learning, prototyping and experimentation in a variety of programming languages.

Provide code snippets, templates, or suggestions for programming-based tasks.

Generated code may contain errors or not work within your context. It may contain security vulnerabilities or inefficient solutions.

Always test and debug generated code thoroughly, understand the logic of the code before use.

Assignment dependent.

21. Interpret Images

Identify patterns, trends or key features within imagery, graphs and visual representations.

Analyse or describe visual data, highlight key information from graphs and charts and interpret visual imagery for meaning and context.

Interpretation may miss important details, misread data, lack wider context and nuance of the subject or research focus.

Verify all observations against the original data, combine AI insights with your own expertise and discuss with peers and seek expert input.

Permitted, but not as a substitute for own judgement.

22. Image Generation

Create visual content such as diagrams, infographics, and conceptual images,

Generate visual materials for presentations, for embedding within essays or academic reports or to supplement case studies.

Images often contain inaccuracies, misleading representation, and many copyright issues.

Seek imagery from trusted open-source or royalty-free sources, ensure proper attribution where used, and only use AI for visual generation if meaningful to the task at hand.

Assignment dependent. In all cases students must understand the underlying data and be able to justify decisions regarding representation.

23. Presentation Generation

Consolidate thinking and use a structure to populate slides and layout for presentation.

Create slides, templates, layouts and structured content either as suggestions or directly into software.

Generated presentations may be generic, lack personal creativity, contain factual errors and fail to suit audience needs.

Use generated presentations as a starting point, review and customise any produced content and adjust visuals, flow and structure to suit the context and meet audience needs.

Permitted to help develop visuals from student's own concepts.

Not permitted to generate core content and argument for the presentation.

24. Proofreading

Improve readability, highlight grammatical issues and suggest corrections.

Review written content for improvements on spelling, grammar, and punctuation.

Accepting changes without review or understanding may unintentionally alter meaning or context and reduce proficiency in effective writing.

Avoid accepting suggestions without first reviewing them and understanding the context in which they're suggested. Maintain awareness of style, tone, and disciplinary conventions.

Permitted in most circumstances but may be forbidden in exercises that directly assess language skills

25. Copy-Editing

Support the refinement of style, clarity and structure in written work.

Suggest rewording, sentence restructuring and improvements in flow based on desired outcomes.

AI may oversimplify complex ideas and there is high risk of losing authorial voice and nuance.

Evaluate and selectively apply suggestions and maintain alignment with disciplinary expectations.

Permitted but students should be wary of letting AI erase their own individual style and 'voice'. May be forbidden in exercises that directly assess language skills

26. Typesetting

Assist in formatting documents according to institutional or publication standards.

Arrange headings, figures, tables, and format references to meet required standards.

Automated formatting may not fully comply, misinterpret expectations and result in errors or altered content.

Avoid automated formatting and seek suggestions to improve or highlighted errors for review. Verify changes against trusted sources.

Permitted

27. Guidance on Citation and Referencing

Support adherence to academic standards by seeking advice on how to correctly reference sources.

Review existing citations and references for errors or inconsistencies and ask for advice on how to align with required standards.

Guidance may be outdated, incorrect or misinterpret style guides. Suggestions may not meet lesser-used standards and could affect your academic credibility.

Verify AI guidance against current style guides and standards, combine advice with personal understanding and seek advice from peers.

Permitted but discouraged, information easily available elsewhere.

28. Generating References and Citation

Automatically create formatted references and citations in a text.

Asking AI to review your work and produce the reference list and generate in-text citations according to the required standard.

References may contain errors, incomplete information, incorrect formatting and incorrectly attribute work to authors.

Author will not have reviewed source material.

Note references as you write and realise a point needs supporting, ideally using a suitable tool but leaving notes to yourself in the text can also work. Do not write a document and then try to add references later.

Not permitted to use AI to attempt to automatically add references to a text based on content.

Use of tools (such as Endnote or Bibtex) which may use Gen AI, for assembling, tracking and formatting of references and citations identified by the student is permitted. The use of automatic and AI tools to check references is also permitted.

29. Guidance on Academic Writing

Help strengthen communications and academic presentation.

Offer advice on structure, clarity, argumentation and tone to improve academic writing.

Guidance may be generic, overlook disciplinary conventions, and relying solely on AI could reduce engagement with materials and limit skill development.

Critically evaluate suggestions to understand context and apply them appropriately. Use guidance to inform and enhance own writing skills and style, not replace.

Permitted but note that the Engineering library may be a better source of such advice (and more friendly!)

30. Personalised Feedback

Provide targeted guidance for improvement and personal development.

Request feedback on drafts, completed activities, seek guidance on academic progress and suggest methods for improvement.

Feedback may be generic and lack subject nuance. Outputs will not appropriately relate to you, your experiences, and your course so will be less effective.

Consider AI feedback critically and as part of a whole rather than the sole source. Use AI feedback to inform reflection and discuss with peers and experts.

Permitted but remember that AI can't make reasoned judgements in the way a person can.

 

 

Last updated on 09/07/2026 17:02