Grammarly got caught selling personalized AI advice from "experts" who never agreed to be experts.
The Signal
Journalist Julia Angwin is suing Grammarly in a class-action lawsuit for using her identity without permission in its "Expert Review" AI feature. The feature claims to offer personalized writing suggestions from real journalists and writers, but those people never consented to their names, likenesses, or reputations being attached to AI-generated feedback.
This isn't a technical glitch. This is a business model choice. Grammarly needed credibility for its AI suggestions, so it borrowed the reputations of actual experts without asking. The company presumably figured the value of social proof (your writing coach is Julia Angwin!) outweighed the legal risk of just taking it.
The lawsuit alleges violations of privacy and publicity rights, specifically laws that prohibit using someone's identity for commercial purposes without consent. Angwin found out she was an "expert" when another journalist, Casey Newton, noticed his own name being used and started asking questions. The Verge uncovered more names in the system.
This touches the central tension in the agent economy: AI systems need human credibility to function, but they're built by companies that treat human identity as just another dataset to scrape. Grammarly isn't selling writing tools anymore. It's selling the borrowed authority of people who built their reputations the old way, through actual work.
The Implication
Watch how this case moves. If Grammarly loses or settles big, it sets precedent that your name, your professional reputation, your identity aren't fair game for AI training or features without explicit consent. Every company building AI agents that claim expertise or personality needs to solve the authentication problem: how do you prove your AI knows what it's talking about without stealing credibility from real people?