Embracing AI in higher education: university policies on generative AI

Artificial intelligence has quickly become a fixture in education. Tools like ChatGPT and Bard are raising both excitement and concern across campuses. Many universities initially feared that generative AI would fuel plagiarism or undermine learning. However, a growing number of institutions have realised that if you can’t beat them, join them. Rather than banning these tools, universities in the UK and US are crafting new policies to integrate AI into teaching and learning. At the same time, they are instituting measures to safeguard academic integrity and privacy. These policies generally encourage responsible use of AI and emphasise transparency and ethics. The goal is to let students and staff benefit from AI’s capabilities without compromising academic standards. We now examine how several universities are approaching generative AI. This overview highlights key guidelines and notable quotes from each institution’s policy.

Existing University policies

University of Oxford

The University of Oxford permits students to use generative AI tools to support their learning. It states:

“You can make use of generative AI tools (e.g. ChatGPT, Claude, Bing Chat and Google Bard) in developing your academic skills to support your studies. Your ongoing critical appraisal of outputs by reviewing them for accuracy will maximise the potential for AI outputs to be a useful additional tool to support you in your studies.”

Oxford’s guidance acknowledges that AI can assist with tasks like brainstorming or practice questions for exams. But it also enforces strict rules to uphold academic integrity. At the same time, it stresses that AI cannot replace human critical thinking. Students must not present AI-generated material as their own work. The policy warns,

“unauthorised use of AI falls under the plagiarism regulations and would be subject to academic penalties in summative assessments.”

The university will treat any unpermitted use of ChatGPT or similar tools on an exam or final paper as academic misconduct. If an instructor or department allows AI assistance on an assignment, students should clearly acknowledge how they used the tool. This balanced approach ensures that learners can experiment with AI. They can do so as long as they remain honest and do not undermine their own skill development.

University of Cambridge

The University of Cambridge has a similar stance that upholds integrity while cautiously embracing AI for learning. Cambridge’s official student guidance makes it plain that students may not use AI-generated content in graded work unless explicitly allowed. The policy pointedly states,

“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.”

That rule means unless an assignment specifically permits AI, students should not rely on it in submitted work. At the same time, Cambridge encourages open discussion about AI use in independent study and formative (ungraded) tasks. The university intentionally worded its new policy to let staff and students engage with these tools more in personal study and practice. This approach invites open dialogue about appropriate use and ethical implications. It helps Cambridge consider what can be gained from AI and what might be lost through over-reliance on such tools. Cambridge is not banning AI outright. Instead, it forbids misuse in coursework but invites students to become informed about AI through guided exploration and discipline-specific norms.

Imperial College London

Imperial College London explicitly promotes effective, ethical, and transparent use of AI while drawing a firm line against misconduct. Imperial’s library guidance notes that generative AI can be a helpful starting point for research or idea generation. It advises students to use these tools critically, verify AI outputs against reliable sources, and remain mindful of the technology’s limitations. However, Imperial is unequivocal that AI must not become a shortcut in assessments. The college directs students to disclose any use of AI in coursework. It also cautions,

“if there is no explicit instruction to use generative AI tools, it would not be considered acceptable to use them to write your assessed work.”

Using AI to write an assignment without permission is simply not acceptable. The policy equates any such attempt with contract cheating — essentially the same as paying someone else to do the assignment. Imperial even warns that departments may conduct “authenticity interviews” with students to verify that submitted work reflects the student’s own understanding. Overall, the college’s policy shows a willingness to integrate AI as a learning aid. At the same time, it demonstrates a commitment to ensure fairness and originality in evaluation.

Harvard University

Harvard University’s approach to AI in the classroom recognises that one size does not fit all. There is no single university-wide ban or blanket rule; instead, Harvard lets each school and instructor set AI usage policies appropriate to their context. Harvard’s Faculty of Arts and Sciences (FAS) has issued broad guidelines rather than a strict edict. For example, in summer 2023 the FAS released recommendations outlining three possible approaches professors might take regarding AI in their courses.

One option is a “maximally restrictive” policy that bans all AI use in coursework. Another is a “fully encouraging” policy allowing AI tools with proper attribution. The third approach is a mixed model permitting some uses but not others. “I don’t think there is a one-size-fits-all course policy here,” one Harvard dean explained. He emphasised that faculty should tailor AI rules to their specific learning objectives.

Harvard’s guidance urges instructors to communicate their AI expectations to students clearly and frequently. It also reminds everyone to follow existing academic integrity standards. In addition, it cautions faculty not to input student work into public AI platforms, for privacy reasons. Harvard is essentially embracing AI by educating instructors and students about it. Each class can then harness AI’s benefits or impose limits as needed, rather than trying to ignore the technology.

Stanford University

Stanford University takes a similar approach. It treats generative AI as an emerging tool to be managed through the Honor Code and course-specific policies. Stanford’s guidance makes it clear that using AI to cheat is unacceptable. Yet it stops short of forbidding AI entirely. By default, if a professor hasn’t explicitly said otherwise, Stanford considers AI assistance akin to help from another person. It must not cross the line into doing the student’s work. The policy states that:

“using generative AI tools to substantially complete an assignment or exam… is not permitted.”

Students must acknowledge any significant AI help they use, just as they would credit a human tutor. They are also encouraged to disclose it “when in doubt.” At the same time, Stanford explicitly gives individual instructors the freedom to allow or disallow AI tools in their courses. Professors might permit AI for brainstorming or drafting, or ban it for certain assignments, depending on their learning goals. Stanford’s overall message is to use AI ethically and transparently. In practice, students should not evade learning by having AI produce their answers. They must always follow the specific guidelines set for each class.

Massachusetts Institute of Technology (MIT)

MIT has approached AI tool use with an emphasis on responsibility, transparency, and data protection. The institute’s guidance encourages students and faculty to use generative AI in an ethical manner — which includes protecting confidential information and maintaining academic honesty. MIT’s policy urges anyone using AI for academic or research purposes to be upfront about it. It explicitly advises:

“You should disclose the use of generative AI tools for all academic, educational, and research-related uses.”

In other words, students should mention if they use an AI assistant to help write code or a lab report. Likewise, instructors should disclose if they use AI to draft an email or quiz questions. MIT also reiterates that users are responsible for checking the accuracy of AI-generated content and for not misusing it. Furthermore, the institute warns its community not to input sensitive personal or institutional data into AI tools like ChatGPT, because those inputs may be retained externally.

While MIT does not flatly prohibit using AI, it ties its use to existing policies on academic integrity and information security. The overarching goal is to integrate AI into MIT’s environment in a principled way. MIT encourages researchers, teachers, and students to experiment and innovate with AI. But it also urges everyone to keep their eyes open to the ethical and practical implications.

Guidance on creating AI policies

A well-crafted policy can harness AI’s benefits for teaching and learning, but it must also guard against academic misconduct. The following guide outlines key considerations and actionable steps to help institutions develop their own AI usage policies.

Key questions to consider

Before drafting an AI policy, university leaders and educators should reflect on fundamental questions about their goals and values. For example:

• What are our educational objectives, and how might AI support or undermine them? (e.g. preserving critical thinking and originality versus using AI for efficiency).
• In which contexts could AI use be beneficial or problematic? Consider differences between formative exercises and summative assessments, and think about whether AI is appropriate in exams, assignments or research projects.
• How will we maintain academic integrity and fairness? Think about what counts as cheating when using AI, and how instructors or tools could detect such misuse.
• What level of transparency do we expect from students and staff? Decide if students must disclose any AI assistance they use. Also consider whether instructors should declare their own use of AI in teaching materials.
• What safeguards are needed for privacy and data security? Plan guidelines so that no one inadvertently shares personal or confidential information with AI tools.

Tailored approaches for AI use

Different courses and disciplines may require different approaches to AI. There is no one-size-fits-all solution, so universities can adopt a range of policy stances depending on context. For instance:

• Prohibit AI in assessments: Some policies completely ban generative AI in coursework where core skills are at stake. For example, a writing-intensive module might forbid tools like ChatGPT for any writing task. This ensures that students produce 100% of the work themselves.
• Allow AI with clear limits: Other policies permit AI as a support tool in specific situations, but with conditions. An instructor might allow students to use AI for brainstorming ideas or for editing drafts. However, students must properly attribute any AI-generated content and refrain from using it in the final submission. (This approach integrates AI as a learning aid while maintaining human oversight.)
• Embrace AI as a learning tool: A more permissive approach encourages students and staff to use AI actively, treating it as a skill to develop. In this case, the policy still requires honesty about AI use. It also insists that students verify any AI-generated information and remain accountable for their own work. The emphasis is on teaching how to use AI responsibly, rather than forbidding it.

Whatever approach a university takes, educators should provide a rationale for it. Explaining why a course prohibits or allows AI helps students understand the policy and follow it. For example, you might highlight that the course bans AI in order to protect certain learning outcomes. Conversely, note if the course allows it in order to foster innovation.

Clear and transparent guidelines

An effective AI policy must be unambiguous so that everyone understands the rules. It should define what counts as “generative AI” and specify which tools or types of assistance fall under its scope. The policy needs to spell out where students may use AI and where they may not. It should also describe what constitutes misuse. For example, the policy can state that presenting AI-generated material as one’s own work is a violation. The document must outline what consequences will follow if such misuse occurs.

The policy should also instruct students on how to acknowledge any allowed AI assistance. For example, it could require adding an “AI usage” statement or a citation to note which tool they used. It is also wise to include a note on data privacy. The policy should caution users never to input sensitive personal or institutional data into public AI tools.

To illustrate, a policy document might include statements like:

“Submitting AI-generated content as your own work will constitute academic misconduct, unless an assessment explicitly permits it.”

“Students may use generative AI tools to support research and drafting. However, they must credit any AI contributions (e.g. in a footnote or appendix), and they remain responsible for the accuracy and integrity of the final submission.”

“If you use AI tools, you are accountable for checking the validity of the AI-generated information. You must ensure it does not contain errors or bias.”

Clear wording like the above leaves little room for confusion. It sets a clear expectation that students use AI only within allowed bounds. This reinforces that honesty and quality are non-negotiable.

Training and ongoing review

Introducing a new AI policy is only the first step. Universities should also educate their communities about the guidelines and the proper use of AI. This can include workshops or tutorials for both staff and students on how to use generative AI ethically and effectively. Instructors, for example, might demonstrate acceptable versus unacceptable uses of AI in their discipline, so students see practical scenarios. Universities should also remind everyone about the risks of AI, such as false information or inherent biases. Users need to stay critical of AI outputs.

Furthermore, building awareness of data privacy is essential. The policy rollout should highlight that no one should paste exam questions, student assignments, or other confidential data into public AI tools. Doing so could breach privacy or copyright. Training users and raising AI literacy ensures that the policy is not just a set of rules. Instead, it becomes part of a broader educational effort.

Finally, universities should treat an AI policy as a living document. University leaders ought to review and update the rules periodically as technology advances and classroom experiences evolve. Gathering feedback from students and faculty can help identify what is working or what needs clarification. With regular re-evaluation, the policy will remain relevant and effective. It will continue to balance innovation in teaching with the fundamental principles of academic integrity.

Wrapping up…

These examples show that leading universities are moving toward an “embrace with caution” philosophy on generative AI. Across the board, institutions agree that completely outlawing AI is neither practical nor pedagogically sound. Instead, they are developing policies that let students and staff leverage AI tools while upholding honesty, critical thinking, and privacy. Common threads include requiring students to acknowledge AI assistance, forbidding unapproved use of AI on graded work, and providing guidance on using AI productively rather than as a crutch. As Yale University’s provost put it, the academic community will “embrace technological tools and harness their power for innovation” — yet always in a safe, responsible manner. Universities are actively adapting to the AI era. They are doing so not by trying to beat the bots at an impossible game of cat-and-mouse, but by joining them in the service of learning. Each institution’s policy is a work in progress, but collectively they mark a significant shift in higher education. Universities are accepting that generative AI is here to stay. The goal now is to integrate it in ways that enrich – rather than erode – academic integrity and educational quality.