Q&A with Superhuman CEO Shishir Mehrotra, who apologized for Grammarly's Expert Review feature, on impersonation, attribution, compensating creators, and more
Today, I'm talking with Shishir Mehrotra, who is CEO of Superhuman — that's the company formerly known as Grammarly, which is still its flagship product.
The VergeNilay Patel
Context & Ripple Effects
Superhuman’s flagship writing product introduced Expert Review as a tool that simulated feedback from recognizable writers without permission, prompting scrutiny over impersonation and creator rights. The company subsequently disabled the feature after backlash, while a separate response offered authors an email-based opt-out process.
The CEO’s apology moves the episode from a product rollback to a public discussion of attribution and creator compensation. It also tests how the company’s broader repositioning around an AI productivity suite—following its rebrand from Grammarly to Superhuman—will handle safeguards in its flagship product.
First-order effects
The apology places Superhuman’s leadership on record over a feature it had already withdrawn, keeping impersonation, attribution, and compensation central to any future version of Expert Review.
Authors whose names or styles were implicated gain more visibility into the company’s stated approach, following the earlier author opt-out process and shutdown.
Second-order effects
AI writing-tool providers face stronger pressure to distinguish general stylistic assistance from features that invoke identifiable creators, and to make consent and attribution policies legible before launch.
For creators, publisher groups, and rights-holders, the episode makes product-level controls and compensation terms more important than post-launch opt-outs alone.
Third-order effects
If similar disputes persist, generative productivity software may shift toward permissioned creator participation, clearer provenance, and narrower use of named-persona features.
The broader governance question is whether voluntary product controls can resolve creator-rights concerns or whether platforms will face demands for more standardized consent and attribution rules.
The trend: This is part of a wider shift from unrestricted AI imitation toward consent, attribution, and control mechanisms for the people whose identity or work informs AI features.
Incredible @reckless.bsky.social interview with Superhuman CEO Shishir Mehrotra, after his Grammarly used our identities without permission. — Nilay does not let up: — www.theverge.com/podcast/8987...
Well it starts off as a relatively friendly interview but oh man does @reckless.bsky.social start slicing and dicing the “Superhuman” nee Grammerly CEO about the company's “impersonation” feature that everyone discovered and quickly loathed. Delicious read. — www.theverge.com/…
It's a fascinating interview with a person basically trying to pull the “we're compensating you with exposure” line and I just wanna point out that because LLMs are bullshit machines Grammarly's “expert” feature was citing completely made-up works. [embedded post]
Here's my interview with Shishir Mehotra, the CEO behind Grammarly's “expert review” feature which attributed writing advice to people - including me lol - without permission. Or, as you will hear us talk about a lot, compensation. www.theverge.com/podcast/8987... [image]
Does Grammarly have lawyers or did they just ask an AI agent if this was legal to do? Even a sycophantic AI agent would have to mention ‘Right of publicity’ laws... [embedded post]
attn to my fellow messy bitches: here's the head of Superhuman getting cross-examined by our fearless leader about the “expert review” feature, which gave weird writing advice with real people's names attached - transcript & video lol www.theverge.com/podcast/8987... [image]
Ohhhh the CEO of Superhuman, née Grammarly, of the “here's bell hooks, Julia Angwin, Nilay Patel etc giving you feedback” feature was running YouTube when they came out with ContentID, which goes a lot to explaining how this happened and how it's going to go: — www.theverge.com…
This interview is wild. The CEO being interviewed cannot conceive of a world where their right to profit does not come first. www.theverge.com/podcast/8987...
finally reading the transcript of the @reckless.bsky.social interview with the grammarly ceo and it's taking foreverrrr bc the ceo speaks like an llm overflowing its context window — i would love a @reckless.bsky.social solo take bc his questions were vastly more informative th…