Harvard Prohibits Use of AI Assistants in Virtual Meetings

The use of artificial intelligence meeting assistants — bots that record and transcribe audio on virtual meeting platforms — will be prohibited at Harvard meetings moving forward, Harvard University Information Technology leadership announced in a Tuesday email.

The email cited potential legal and data security risks as motivations for the ban, in addition to concerns that AI assistants “have the potential to stifle conversation and open inquiry.” In particular, the updated guidelines warned that meeting data could be exposed to third parties or used by companies when training future AI models.

The ban will not, however, apply to assistants with which the University maintains contractual enterprise agreements.

HUIT spokesperson Timothy J. Bailey wrote in a statement that HUIT had worked with Harvard administrators and its Office of General Counsel to develop the new guidelines after receiving “community requests for more information on AI meeting assistants.”

Bailey added that the University was conducting a pilot program to assess the viability of Zoom’s meeting assistant, Zoom AI Companion.

The updated guidelines come just one week after a Faculty of Arts and Sciences committee recommended the adoption of Chatham House rules, which would bar students and instructors from attributing comments made in classroom discussions to specific speakers.

Computer Science lecturer Christopher A. Thorpe ’97-’98 wrote in a statement that he believed the updated guidelines were “a thoughtful, practical balance of making these innovative tools available—while keeping longstanding commitments to data privacy and the comfort of meeting participants.” Thorpe teaches COMPSCI 1060: “Software Engineering with Generative AI.”

“I also think the Zoom AI Companion pilot is a great idea as a starting point,” he added.

HUIT’s updated guidelines acknowledge the role of AI note-taking assistants in aiding attendees with accessibility concerns and instructed individuals who required AI accommodations to reach out to their accommodations coordinator “to discuss options.”

The guidelines urged Harvard affiliates to avoid using AI assistants in sensitive meetings, review and correct AI-generated summaries, and delete past transcripts once no longer needed.

Subir Majumder, an Electrical Engineering postdoctoral researcher, said he felt AI assistants had become “intrusive” in his meetings prior to the updated guidelines.

“During certain meetings, one of my colleagues also had some automatic note-taker who automatically joined in. My boss then said, ‘Who is this guy?’” Majumder said. “This creates a little bit of miscommunications.”

However, A. Rhys Greenland ’27 said he felt the new policy had a minimal impact on him as an undergraduate student.

“It’s not a big deal,” he said.

—Staff writer Xinni (Sunshine) Chen can be reached at sunshine.chen@thecrimson.com. Follow her on X @sunshine_cxn.

—Staff Writer Danielle J. Im can be reached at danielle.im@thecrimson.com.