Over 350 staff members at the University of Edinburgh have signed an open letter demanding that the institution does not renew its contract with OpenAI, the creators of ChatGPT. The protest, which has gained significant attention in higher education circles, centers on ethical, security, and alignment concerns with the university's principles. As the contract approaches its end, this movement highlights growing tensions around artificial intelligence (AI) partnerships in academia, particularly in light of OpenAI's recent collaboration with the US military.

The University of Edinburgh's Edinburgh Language Model (ELM) platform provides staff and students with access to various large language models (LLMs), including those from OpenAI. ELM serves as a secure gateway to generative AI, emphasizing responsible use through guidelines that require users to agree to ethical standards before access. However, critics argue that including OpenAI's proprietary models contradicts these very standards.

The Open Letter: A Collective Call for Change

The open letter, addressed to the EDINA leadership team responsible for the OpenAI contract, meticulously outlines why continuing the partnership is untenable. Signatories include prominent AI experts such as Adam Lopez, Reader in the School of Informatics; Zeerak Talat, Chancellor's Fellow in Responsible Machine Learning; and Shannon Vallor, Baillie Gifford Chair in the Ethics of Data and AI. Other notable names encompass researchers, lecturers, and administrative staff from across disciplines, demonstrating broad community support.

James Galbraith, a postdoctoral research associate in the School of Biological Sciences, articulated a core sentiment: "The central issue is that contracting OpenAI to provide LLMs to staff and students does not follow the university’s AI policies, in particular the labour rights issues, the impact their data centres are having on the communities they have been built in, and their contracts with the US military." The letter urges a shift exclusively to open-weight models hosted locally within ELM, allowing better monitoring of energy use and alignment with ethical procurement.

The document is available for public viewing and signing via Google Forms, reflecting transparency in the protesters' approach.

Background on the University of Edinburgh's AI Initiatives

ELM, launched as the university's AI innovation platform, aims to offer safer access to generative AI tools. It integrates multiple LLMs, with OpenAI's models among them, paid centrally via API on a token basis to reduce costs significantly—up to 90% in some cases. The platform mandates that users read and accept AI guidelines, covering data protection, ethical use, and risks like hallucinations or bias.

The University of Edinburgh, a global leader in informatics and AI research, has invested heavily in this infrastructure. Information Services Group (ISG) collaborates with EDINA to ensure responsible exploration of emerging technologies. Guidelines for staff and students emphasize limitations of generative AI, such as potential inaccuracies, and promote its use as a supportive tool rather than a replacement for critical thinking. More details on ELM can be found on the official university page.

Core Concerns: OpenAI's US Military Ties

A pivotal trigger for the protest is OpenAI's February 2026 agreement with the US Department of War (formerly Defense), enabling deployment of its AI models in classified networks. Announced shortly after the Trump administration blacklisted rival Anthropic for refusing unrestricted military use, the deal faced immediate backlash for potential involvement in autonomous weapons and surveillance.

OpenAI's official statement details "red lines": no mass domestic surveillance of US persons, no directing autonomous weapons (requiring human oversight), and no high-stakes automated decisions like social credit systems. Despite these safeguards, protesters view it as enabling semi-autonomous weaponry, conflicting with UN calls for bans on lethal autonomous weapons systems (LAWS). The letter also cites OpenAI's services to US Immigration and Customs Enforcement (ICE), accused of paramilitary actions.

For full details, see OpenAI's agreement post.

Safety and Security Risks Highlighted

The letter documents OpenAI's poor safety record: multiple court cases involving harm, including suicides linked to ChatGPT interactions. Security is another flashpoint, with OpenAI reporting the highest number of data breaches among LLM providers and a low security score. Protesters contrast this with alternatives like Mistral, which prioritize privacy by design.

Responsible access is questioned since OpenAI models run remotely, evading local energy monitoring—a key university principle. OpenAI's lobbying against AI regulation, including $100 million in Super PAC funding and donations to political figures, further erodes trust.

Photo by 𝕡𝕒𝕨𝕤 𝕒𝕟𝕕 𝕡𝕣𝕚𝕟𝕥𝕤 on Unsplash

Ethical and Environmental Misalignments

OpenAI's procurement practices violate the university's Responsible Procurement Policy, which demands carbon emission targets, fair work commitments, and human rights alignment. Data centers' massive climate impact remains opaque, conflicting with net-zero goals. Labor exploitation is stark: Kenyan workers earned less than $2/hour moderating traumatic content, described as "modern-day slavery" in open letters.

Bias in models perpetuates discrimination, with studies showing unprecedented discriminatory attitudes toward Black Americans and place-based inequalities. Transparency lags, as proprietary nature hinders accountability in research and education.

The University's Measured Response

Gavin McLachlan, Vice-Principal, Chief Information Officer, and Librarian, responded thoughtfully: "The university aims to provide all students and staff safer access to AI tools and technology in a way that aligns with our values. We welcome the opportunity to engage with our community... and plan to discuss the concerns directly with the authors." He highlighted existing AI training, guidelines, and secure platforms protecting data privacy.

The administration positions ELM as offering choice among LLMs in a managed framework, supporting diverse disciplines without compromising security.

OpenAI's Defense Against Criticisms

An OpenAI spokesperson countered: "This letter makes misleading claims about OpenAI. We’re focused on building AI that is safe, useful, and benefits as many people as possible." They emphasize heavy safety investments and global government collaborations for responsible deployment.

Broader Implications for UK Higher Education

This protest echoes global debates on AI ethics in academia. The University of Oxford partners with OpenAI for education-focused ChatGPT and grants, while Manchester uses Microsoft Copilot. In the UK, concerns over military AI ties, job displacement, and ethics are mounting, with calls for open-source alternatives.

Stakeholders like the Centre for Technomoral Futures at Edinburgh underscore the need for responsive policies. As AI integrates deeper into teaching and research, universities must balance innovation with principles—a challenge for institutions worldwide.

Stakeholder Perspectives and Expert Insights

• AI Experts: Signatories like Shannon Vallor advocate for ethical oversight, warning of societal risks.
• Students: Concerns over skill erosion from over-reliance on LLMs.
• Administrators: Value cost savings and managed access but face pressure to diversify providers.
• Industry: OpenAI pushes accessibility; critics push local, open models like Llama.

Future Outlook: Navigating AI in Academia

The University of Edinburgh is benchmarking more local LLMs, potentially phasing out OpenAI. This could set a precedent, encouraging procurement based on ethics audits. Actionable steps include prioritizing open-weight models, enhancing transparency reporting, and stakeholder consultations.

For higher education professionals, this underscores the importance of aligning tech partnerships with institutional values, fostering innovation responsibly amid rapid AI evolution.

For comprehensive coverage, refer to the Times Higher Education article.