Durham University staff permitted to use GenAI in summative marking

By Milly Moore and Abigail Tobias

Durham University expands ambition for AI Research and growth while explicitly confirming the permissibility of staff using GenAI tools. These tools include Microsoft Copilot for research and marking and expanding AI research.

Palatinate has seen a University Senate Generative AI policy paper stating that staff are permitted to use GenAI in summative marking. The document notes: “Staff may explore the potential to use GenAI to assist but not replace academic judgement. The use of GenAI in marking processes must be explicitly identified.”

However, AI policy for staff is often determined at departmental level, resulting in variation across the University. Palatinate has seen documents from Accounting stating that staff are allowed to use GenAI in summative marking in line with University policy, whereas Geography operates a strict ban.

A Durham University spokesperson said: “We recognise the potential of Generative AI to support our staff and students, including in preparing students for a workplace where Generative AI is increasingly present. We are also conscious of risks, and do not compromise on academic standards, so are approaching its adoption with caution.

“We have a clear University framework on how Generative AI may and may not be used in assessment. This is reviewed every year. It includes that use in marking can be explored, in an open and transparent manner. Departments must base their local policy on the University framework and in all circumstances, departments must clearly explain to students what is and is not allowed.”

“We have a clear University framework on how Generative AI may and may not be used in assessment. This is reviewed every year. It includes that use in marking can be explored, in an open and transparent manner” Durham University Spokesperson.

This change was brought about following information about the University’s largest risk being financial instability, which they plan to tackle through cutting staff costs – a detail highlighted by the Palatinate (link strategic risk article) after the publishing of Durham’s most recent annual financial report.

This annual report claims that “a greater level of ambition will be needed to meet the university’s targets” for research income, and that incorporating developing uses of AI is part of that ambition. The University clarifies its goal stating, “GenAI has a great potential to transform teaching, learning, assessment, ways of working, improve access to information, create content, analyse data and enhance workplace efficiencies.

As a university, we are committed to empowering our staff and students to use these technologies in an increasingly AI-enabled world.”

This is in reference to the University’s announcement that Copilot Chat has been made available to all users in the University, and that “Microsoft 365 Copilot will undergo phased implementation on a use case by use case basis.”

This means that according to departmental preferences, staff are permitted to use Copilot alongside their own professional judgement to mark student’s work.

The University ensures that “Copilot Chat comes with Microsoft’s Enterprise Data Protection standards, meaning Microsoft have no eyes access to the information that is provided to Copilot Chat and data is not used to train any language models.”

However, it does acknowledge the other risks regarding the ethical, environmental, privacy and data concerns, the potential for bias, and the inaccuracy and misinterpretation of information. Therefore, the university states that “GenAI must therefore be used responsibly, and individuals are ultimately accountable for using the outputs from GenAI tools.”

The Government UK website states the regulatory stance in the UK, clarified by Ofqual, is that AI used as the sole or primary marker for regulated, high-stakes assessments is not compliant with regulations. As remarked, “currently, marking for high stakes qualifications is underpinned by the application of academic judgement by a trained expert.”

Despite this, there is increased risk of dangers with what researchers call “capability unpredictability”, which describes the idea that LLM’s (Large Language Models) capabilities cannot be fully anticipated, even by the model’s creators, until its behaviour on certain input it observed. Microsoft copilot relies on LLMs which can produce inconsistent, inaccurate, or “hallucinated” answers.

The UK Government website explains; “The motivations for using AI in a marking process can include improving the consistency of marking, increasing the perceived fairness of marking and – most commonly – improving efficiency by reducing marking time, marking cost and the number of human experts required.”

The allowance of Copilot is an endorsement of the Russell Group principles on the use of the generative AI tools in education, namely:

–              “Universities will support students and staff to become AI-literate.

–              Staff should be equipped to support students to use generative AI tools effectively and appropriately in their learning experience.

–              Universities will adapt teaching and assessment to incorporate the ethical use of generative AI and support equal access.

–              Universities will ensure academic rigour and integrity is upheld.

–              Universities will work collaboratively to share best practice as the technology and its application in education evolves.”

This aligns with the University’s recent creation of a new University Research Institute to provide a governance framework for the Leverhulme Centre of Algorithmic Life (CAL). This corresponds with the aim of reducing staff-costs as the University reassures that “no additional staff costs are envisaged by the conferral of URI status other than the responsibility allowance paid to URI Directors. If, as anticipated, CAL generates additional grant funding beyond the LT grant” then there may be additional roles in an administrative capacity.

This marks one of the University’s achievements in pioneering AI research in the wake of the UK Government’s September 2025 announcement that North East England will become an AI Growth Zone.

Image Credit: James Tillotson