Grammarly Class Action: AI Expert Review Used Authors' Identities Without Consent (2026)

Grammarly Class Action: AI Expert Review Used Authors' Identities Without Consent (2026)

Grammarly is facing a class action lawsuit over its now-shutdown AI Expert Review feature, which falsely attributed editing suggestions to real authors like Toni Morrison, Stephen Pinker, and Noam Chomsky — without their knowledge or consent. The feature, deactivated on March 11, 2026, mimicked the writing styles and reputations of prominent literary figures, creating a deceptive illusion of expert endorsement. According to factually.co, this practice constitutes digital impersonation and violates core principles of AI ethics and authorial rights.

How Grammarly’s AI Expert Review Feature Worked

The AI Expert Review tool analyzed user text and generated feedback styled to resemble the voice of acclaimed writers. Internal documents, leaked to investigative outlets, revealed the system was trained on publicly available works to replicate tone, syntax, and rhetorical patterns. Crucially, it appended fake author attributions like "As Toni Morrison once wrote" directly in user-facing outputs. No consent was sought from the individuals whose work was used.

Training Data Sources and Lack of Consent

Grammarly reportedly scraped millions of published books, essays, and articles to train its model. While public domain content is often used in AI training, the direct use of names and stylistic signatures in consumer interfaces crosses into uncharted legal territory. Authors only discovered the misuse after users shared feedback bearing their names — sparking outrage on social media and literary forums.

Why This Is Different from Standard AI Training

Unlike OpenAI or Anthropic, which avoid naming individuals in outputs, Grammarly explicitly invoked real identities to boost perceived credibility. This turns passive data use into active impersonation, blurring the line between inspiration and fraud. Legal scholars warn this could set a precedent for how AI companies must handle public figures’ creative identity.

Legal Precedents for AI Identity Theft and Data Privacy

The class action, filed in the Northern District of California, seeks injunctive relief and statutory damages under state privacy and right-of-publicity laws. It cites California’s CCPA and the EU’s AI Act as emerging frameworks that may apply. Courts have yet to rule on whether AI-generated impersonation qualifies as identity theft — but this case may force the issue.

Similar Cases in Tech: OpenAI, Anthropic, and Beyond

While OpenAI and Anthropic have faced scrutiny over training data, neither has directly attributed outputs to real people. Grammarly’s approach is uniquely risky. In 2025, a similar case against a resume-building AI tool resulted in a $12M settlement for unauthorized use of LinkedIn profiles. Legal analysts predict this lawsuit could follow suit.

What Users and Authors Can Do Now

If you used Grammarly’s Expert Review feature between January 2025 and March 2026, you may be eligible to join the class action. Affected authors are encouraged to document instances where their name or style was misused. Visit the official lawsuit portal at courtfilings.us/grammarly-ai-2026 for updates and claim forms.

Why This Case Could Redefine AI Ethics in 2026

As AI tools become embedded in daily communication, trust is the most valuable currency. Grammarly’s misstep didn’t just break rules — it eroded user confidence in AI as a neutral assistant. Regulatory bodies in the U.S. and EU are accelerating proposals requiring transparent AI attribution. This lawsuit may become the benchmark for enforcing accountability in generative AI.

Grammarly’s decision to shut down the feature may have been reactive, but the damage to its brand reputation is likely permanent. The real question isn’t whether AI can mimic human voices — it’s whether companies should be allowed to pretend they are human at all.