University of California signs systemwide OpenAI deal as faculty urge caution

On 21 October, the faculty leaders at the top of the University of California's Academic Senate learned that a deal was pending. It would put ChatGPT in the hands of hundreds of thousands of students, faculty and staff. Nine days later, the Academic Council, the executive body made up of Senate leaders, wrote to UC President James B. Milliken. It did not oppose AI, the letter said, but it had questions, and it asked the university to delay. The university went ahead anyway.
What the deal does
The systemwide agreement with OpenAI covers all 10 campuses and will roll out in the coming weeks. It includes ChatGPT Edu and Codex, OpenAI's coding tool, with basic models provided centrally and discounted access to more advanced models for campuses and users who want them. Academic programmes and operations are in scope. Hospitals and clinics are not. The deal expands a looser 2024 arrangement under which each campus bought its own ChatGPT Edu licences and credits and paid its own way, and it places UC alongside Cal State and USC as California universities partnering with OpenAI.
The case for centralising is a serious one. Milliken told faculty leaders the purchase "will help ensure equitable access to these widely used tools while providing stronger data protections", and said it had buy in from chancellors. UC San Diego Chancellor Pradeep Khosla put it plainly: "As an educator, my view is I should give equitable access to all my faculty, students and staff." Miguel Craven, a UC Davis graduate student and student regent, described the gap the deal could narrow. Some students have paid for premium models for a long time, he said, while others "can't afford to purchase those models. So the free resources are what they rely on, and it has led to an equity gap." With an April survey by the Lumina Foundation and Gallup finding 57% of college students used AI for coursework at least weekly, that gap is not hypothetical.
Where the trust breaks
The dispute is less about AI than about who decides. "Foundational decisions about UC's educational environment are being made without sufficient involvement of its educators," said Susannah Scott, Academic Senate chair and a UC Santa Barbara chemistry professor, adding that the Council's concerns remained unresolved. Milliken's answer is that offering the tools "in no way mandates their use by students or faculty, nor does it establish academic policy."
Greg Leazer, an associate professor of information studies at UCLA, rejects that framing outright. "The University of California's purchase of OpenAI licenses for all of its students, faculty and staff is a significant new academic policy masquerading as a procurement issue," he said. He is right about one thing, whatever one makes of his view that ChatGPT "reduces learning". When a tool lands on every desk, defaults shift, and defaults are policy. The evidence will not settle the argument either. Research on whether AI helps college students learn is limited: a Harvard study found physics students learned more with a custom AI tutor than in class, while other studies found little benefit, and cheating concerns are widespread.
Khosla said AI "conjures up images of euphoria and fear simultaneously" and that both camps have "legitimate reasons" for their views. That is precisely why the process matters. Access is a technical decision. How the tool is used in a seminar, an assessment or a research lab is a professional one.
What this means for L&D
Any large organisation rolling out AI faces the UC dilemma. Central procurement brings equity, better data protection and consistency. It also invites the people who run the work to treat the tool as something done to them. Saying a rollout does not mandate use is not the same as having norms for use, and in the gap between the two, trust erodes.
The fix is sequencing and ownership. Pair access with norms written by practitioners: the teachers, engineers, analysts and managers who know where the tool helps and where it undermines the craft. Give those groups a formal role before launch, not a letter after it. This is the work we focus on in our AI Governance programmes, because a licence count is easy to report and a shared understanding of good use is what makes it worth anything.
Sources:
News and insights for innovation, digital transformation, future of work and L&D leaders.
Stay ahead of learning and development, corporate innovation and digital transformation news. Plus the future of work. For leaders in AU, NZ, HK, SG, the US, the UK and Canada.





