I have been at the University of Bath for twenty years, and in that time few developments have presented a challenge, and an opportunity, of the scale now posed by Generative AI (GenAI). Unlike previous waves of educational technology, tools such as ChatGPT are not peripheral to the academic environment, they are already reshaping how students learn, write, and engage with assessment. They raise important questions about academic integrity, the design of meaningful assessments, and the broader purposes of higher education (HE). It is in response to these questions that GEOG, the Generative Artificial Intelligence in Education Oversight Group, was established. As Co-Chair, alongside Christopher Bonfield, Director of the Centre for Learning and Teaching, I welcome this opportunity to introduce the group formally to colleagues across the University, to set out what GEOG is, what it has been working on, and the direction in which it is heading.
What is GEOG?
GEOG is a formal sub-group of the University’s Education Advisory Board (EAB), established to provide institutional leadership on the use of GenAI in relation to teaching, learning, and assessment. Research, postgraduate research, and the operational use of GenAI tools by staff in their day-to-day work fall outside its remit. The group monitors progress against the Senate-approved Russell Group GenAI Principles, a shared framework across leading UK universities, and reports formally to the EAB. GEOG’s remit encompasses the following core responsibilities:
- Identifying and communicating the risks, benefits, and opportunities of GenAI in relation to teaching, learning, and assessment.
- Coordinating GenAI-related activity across the institution to ensure a coherent and joined-up approach.
- Advising on priorities and recommending policy for the effective, ethical, and safe use of GenAI by students, academic staff, and professional services.
- Overseeing the challenges GenAI poses to academic integrity and advising on short, medium, and long-term responses.
- Supporting the development, implementation, and review of guidelines, resources, and policies relating to GenAI.
- Monitoring national and international developments and collaborating with peer institutions on shared challenges and emerging best practice.
GEOG brings together colleagues from across the institution, including the Centre for Learning and Teaching (CLT), Academic Registry, the Library, the Skills Centre, as well as representation from various academic departments from across the four Faculties/School. This cross-institutional composition reflects a considered view that the challenges posed by GenAI are not confined to individual departments, and that an effective institutional response requires engagement from across the full breadth of the University’s academic and professional services communities.
What GEOG Has Been Working On
GEOG has sought to move beyond policy discussion towards substantive, practical engagement with GenAI. In partnership with Amazon Web Services (AWS) and a third-party educational technology partner, EDT, the University has a dedicated AWS instance to install Amazon Bedrock, a platform for accessing and evaluating large language models. This means any LLM running on Bedrock is ringfenced within our own environment; data is not shared with any third-party.
The CLT has been exploring options, built on this secure foundation, for the development of in-house GenAI tools. The objective is to identify solutions that are genuinely fit for purpose (e.g. they help staff to solve real-world challenges, and can be evidence based), and appropriately safeguarded in terms of data integrity, accuracy, and institutional values. Two agents are in development: (1) a feedback coach for staff that checks alignment of feedback to marking criteria, and suggests improvements for clarity; (2) an exam preparation tool for students. In both instances, the human remains in the driving seat: AI supplements and supports, but does not replace human judgement or critical thinking.
To date, CLT has produced an extensive guidance for staff; and the Skills Centre has produced resources for students on effective and ethical considerations in GenAI usage. These are available now and colleagues are encouraged to engage with them. The CLT also has a range of self-paced guides, and can offer 1:2:1 support and bespoke training.
The Move to a Two-Lane Approach to Assessment
The most significant contribution GEOG has made to date concerns the shift in the University’s approach to assessment in the context of GenAI. For the past couple of years, in line with much of the sector, we have operated an A/B/C categorisation system: Category A designating assessments in which no use of GenAI was permitted; Category B designating assessments in which GenAI was permitted for specific, defined purposes; and Category C for those assessments in which GenAI use was integral to the task itself. The system provided a workable initial framework and was widely adopted across the sector. However, in practice it has given rise to some inconsistency and confusion, particularly in relation to Category A and B assessments, where the boundaries of permissible use have proved difficult to define and communicate clearly to both staff and students.
In August 2025, GEOG proposed a transition to a two-lane framework. This was approved by EQSC on 2 July 2025. This year has operated as a transition year – and myself and the CLT have been running workshops with Departments to help explore what this means within the context of their own disciplines and assessments. From AY 26/7, A/B/C will be “shut down” and be replaced formally with two-lane. This framework establishes a simpler and more principled distinction:
- Lane 1 — Closed assessment: assessments in which the use of GenAI tools is not possible, with security maintained through the conditions of the assessment itself — for example, invigilated in-person examinations, or other in-person tasks assessments such as oral assessments and Vivas.
- Lane 2 — Open assessment: GenAI usage is permitted. These are assessments in which integrity and validity are ensured through effective assessment design, rather than through restrictions on tool use. The emphasis shifts from the policing of permitted tools towards the design of assessments that are genuinely meaningful, personally situated, and resistant to outputs that rely on superficial GenAI use.
This framework should not be seen as a relaxation of standards with respect to academic integrity. Rather, it reflects a more considered and sustainable position: that the most effective response to the challenges GenAI poses to assessment is not the imposition of increasingly contested restrictions on tool use, but the design of assessments that are inherently worth completing and that generate meaningful evidence of student learning.
More communications are planned for staff and students on this (stay tuned to this blog series for forthcoming posts!), but guidance on what the two-lane framework means in practice for assessment design has been shared with Associate Deans for Education to disseminate to their respective areas, and the CLT has produced a FAQ page to assist.
Conclusion
GenAI presents the sector with challenges that are unlikely to diminish, and opportunities that would be unwise to overlook. The institutions best placed to serve their students will be those that engage with these developments rigorously, honestly, and with a clear sense of educational purpose. That is the commitment that GEOG represents. I hope this introduction provides colleagues with a clearer understanding of the group and its work. Further updates will be shared on a regular basis, and colleagues with questions or who wish to contribute to the group’s work are encouraged to make contact.
Blog Post By: James Fern, Department for Health.
James is co chair of the Generative Oversight Committee Group. If you would like any further information please get in touch with James via email jf257@bath.ac.uk
