Grammarly's "Expert Review" Exploited Authors Without Permission
Grammarly suspended its "expert review" service last week after technology journalist Julia Angwin filed a federal class-action lawsuit alleging the company used authors' names and likenesses without consent to market writing feedback generated by AI bots.
The service, which cost up to $30 per month, promised subscribers that established writers-including Stephen King and Neil deGrasse Tyson-would review their work and offer personalized feedback. None of that happened.
How the Service Actually Worked
Users uploaded documents to Grammarly's system. AI bots analyzed the writing and generated suggestions based on the authors' published works. The feedback appeared to come directly from the cited experts, with disclaimers buried deep in the fine print stating suggestions were merely "inspired by" the experts and didn't indicate endorsement or affiliation.
The roster included deceased writers. Grammarly listed astronomer Carl Sagan (died 1996) and grammarian William Strunk Jr. (died 1946) as available reviewers.
The company did not publish a comprehensive list of whose names it used or seek permission from any of them.
The Legal Challenge
Angwin, who has worked at the Wall Street Journal and ProPublica, discovered Grammarly was exploiting her name through a report by Casey Newton of Platformer. She said the AI-generated edits proposed under her name were poor quality, damaging both her livelihood and reputation.
Her lawsuit cites California and New York laws prohibiting commercial use of anyone's name or likeness without consent. Angwin's attorney, Peter Romer-Friedman, said more than 100 writers have contacted him about joining the case since it was filed, with the total potentially reaching thousands.
Grammarly's CEO Shishir Mehrotra said the company "believes this feature missed the mark" but called the lawsuit claims "without merit" and said the company would "strongly defend against them."
Part of a Broader Pattern
This isn't the first time AI companies have cut corners. Legal briefs with citations to nonexistent cases, medical diagnoses recommending dangerous treatments, and news articles published without human bylines have all exposed the risks of deploying AI without adequate oversight.
Grammarly's move comes as authors and artists pursue multiple lawsuits against AI companies over training data. Courts have not yet settled where fair use ends and copyright infringement begins.
Tech journalist Kara Swisher called Grammarly's approach "rapacious information and identity theft."
What Happens Next
Grammarly initially offered writers the chance to opt out by sending an email, but the company never disclosed which names it was using. The company told one inquirer only that it sourced "expert" data from third-party AI bots and wouldn't provide additional comment.
Romer-Friedman said obtaining the full roster of exploited writers would be his first priority if the case goes to discovery. Superhuman, Grammarly's parent company, has not yet responded to the lawsuit in court.
Learn more about AI for Writers and the broader issues surrounding Generative AI and LLM training practices.
Your membership also unlocks:
