FAS Plans to Grant Access to Anthropic’s Claude, Phase Out ChatGPT Edu

Harvard’s Faculty of Arts and Sciences will add Anthropic’s Claude to the suite of artificial intelligence platforms available to affiliates while discontinuing access to ChatGPT Edu, FAS Senior Advisor on Artificial Intelligence Christopher W. Stubbs confirmed last week.

Stubbs, a Physics professor, said the FAS plans to adopt Anthropic’s Claude Code toolkit, though he did not confirm a timeline for the rollout. He declined to share which specific tools would be offered to affiliates, but said access would likely be granted on a course-by-course basis, with faculty requesting the program.

Dean of the School of Engineering and Applied Sciences David C. Parkes acknowledged the plan at an event last Thursday on the use of agentic AI in research, which focused largely on Anthropic’s technology.

“The short answer is, we’ve been providing ChatGPT OpenAI technology, and we will be providing Anthropic technology,” Parkes said.

“It’s not that you necessarily should just provide one tool, but tooling generally is important, and we want to keep doing that, and we also want to make sure that our students are aware of how to use these tools,” he added.

The shift comes as the FAS winds down a pilot program that underwrote OpenAI enterprise accounts for all affiliates. After June 2026, access to ChatGPT Edu will require “administrative and budgetary approval,” according to the FAS website.

The FAS also offers affiliates access to Google’s Gemini — another generative AI platform — through an existing institutional agreement between Google and Harvard. The school hosted workshops on the tool in January and March.

FAS spokesperson James M. Chisholm wrote in an email that affiliates’ access to Gemini was not changing.

Stubbs attributed the end of the OpenAI program to financial considerations and low student uptake.

“The uptake among undergraduates was far less than we anticipated,” he said. He attributed the low usage to unfounded student concerns about being caught cheating.

“The anecdotal evidence that we get back is that, ‘Oh, the University did this so that students use those accounts so that we can catch people cheating.’ That was just emphatically not the case, right? It was really an attempt to have equity of access across the student population,” he added.

Stubbs said he did not expect the shift to significantly affect affiliates, pointing to the range of tools already available through Harvard credentials.

“I think people are going to find that the tools that are available to people with g.harvard.edu authentication — access through Gemini and the like — are very much functionally similar to what OpenAI is,” he said.

Faculty appeared largely unconcerned by the change.

“Moving on from ChatGPT is fine with me,” Organismic and Evolutionary Biology professor Bence P. Ölveczky wrote in an email.

Kevin A. Rader, a senior preceptor in Statistics, wrote that “losing the EDU version of ChatGPT is a little bit of a bummer,” but added that “the shift will not affect my day-to-day” because he primarily uses a sandbox offering several AI platforms.

Chisholm said Thursday that affiliates should not expect Harvard to commit to a single platform long-term.

“Given how rapidly evolving the space is, and that you have new technologies and applications and use cases emerging all the time — let alone brand new companies — this is going to be something that is continually evaluated by the FAS,” he said.

—Staff writer Sebastian B. Connolly can be reached at [email protected] or on Signal at sbc.23. Follow him on X @SebastianC4784.

—Staff writer Summer E. Rose can be reached at [email protected] or on Signal at ser.85. Follow her on X @summerellenrose.