The University of Michigan said it asked one of its vendors to stop work, following an offer on social media to sell student data to train artificial intelligence.
“Student data was not and has never been for sale by the University of Michigan,” spokesperson Colleen Mastony said in a statement to Inside Higher Ed. “The [message] in question was sent out by a new third party vendor that shared inaccurate information and has since been asked to halt their work.”
Last Thursday, a Google employee posted on the social media site X a screenshot of a sponsored message received on LinkedIn. The message, from an unknown company, said the University of Michigan was “licensing academic speech data and student papers” that could “be very useful for training or tuning LLMs,” or large language models, which are used to train artificial intelligence.
The message said the potential training materials included 829 student papers, 65 speech events and 85 hours of audio recordings. It set the license fee for the complete set of data as $25,000.
Mastony said the papers and recordings referenced in the post had been voluntarily supplied by students, with signed consent, from two research studies that spanned from 1997 to 2000 and 2006 to 2007. The content did not include names or personal data.
The papers and recordings have “long been available for free to academics” and used as a tool to improve writing and “articulation in education,” Mastony said.
