The company formerly known as Grammarly has a new name, a new CEO, and a very uncomfortable question to answer: did its AI impersonate a real person?
Superhuman CEO Shishir Mehrotra sat for an interview in which he was directly confronted about the company's AI tools replicating a journalist's writing voice without consent. Mehrotra, who previously served as chief product officer at YouTube and currently sits on Spotify's board of directors, now leads the rebranded company. Grammarly itself remains the flagship product under the Superhuman umbrella.
The impersonation allegation hits differently than the usual AI ethics debate. This isn't about generating generic text. It's about an AI trained on enough of someone's writing to convincingly mimic their personal style. Grammarly has collected writing samples from over 30 million users for years. When that data trains AI models capable of producing text that sounds like a specific individual, "writing assistance" starts looking a lot more like identity borrowing.
Mehrotra's decision to address the controversy face-to-face rather than through a PR statement is unusual for an AI company CEO. Most dodge these conversations entirely. But the rebrand itself signals bigger ambitions than catching typos and suggesting synonyms. Superhuman is positioning for a broader AI play, and broader ambitions attract broader scrutiny.
For anyone who's been typing into Grammarly for years, this raises a practical question worth investigating: what exactly happens to the text you submit? The company's privacy policy and terms of service are worth a careful read, particularly around training data and model improvement clauses.
The name change from Grammarly to Superhuman also marks a notable shift in how AI writing companies see themselves. Grammar checking was the entry point. Generating content in your voice appears to be the destination.