Grammarly’s CEO defends putting AI editorial suggestions into the voices of real writers (while noting it didn’t work very well)

Sometimes you just get lucky with the editorial calendar. About a month ago, Superhuman CEO Shishir Mehrotra agreed to be a guest on Decoder, the podcast from The Verge hosted by editor-in-chief Nilay Patel. Superhuman is the company formerly known as Grammarly, which is now just one of its AI-focused productivity tools, and Mehrotra and Patel would have already had plenty to talk about. But then journalists started noticing a little-known Grammarly feature called "Expert Review," which offered users suggested improvements to whatever they were writing. But the proposed changes were framed as coming from real writing experts—journalists, novelists, academics, and more. It wasn't AI suggesting a plot twist—it was Stephen King. It wasn't AI proposing a cleaner way to explain science—it was Carl Sagan. And it wasn't AI pointing out a way to improve your tech column—it was Nilay Patel.

Writers, as one might expect, were not thrilled to see their names being used to lend credibility to AI editing suggestions. (Even me!) Technology journalist Julia Angwin filed a class-action lawsuit against Grammarly, seeking damages. By that point, Mehrotra had announced the demise of "Expert Review" after eight months, saying he wanted "to apologize and acknowledge that we’ll rethink our approach going forward."

Give Mehrotra credit for showing up for his interview with Patel anyway, which was released this morning. After a brief chat about Superhuman's product strategy overall, Patel dove straight into the dispute:

"You do not have our permission to use our names to do this. You had little check marks next to the name that indicated it was somehow official. People did not like this, I did not like this, and you removed the feature. Tell me about the decision to launch this feature with names you didn't have permission for and the decision to unlaunch the feature."

Mehrotra apologized again ("I'm really sorry about that"), admitting that the feature was a mistake. He explained that the idea behind "Expert Review" was to make the AI suggestions more relatable and trustworthy. The company had partnered with a small group of writers to create the suggestions, and they used the writers' names to personalize the feedback. However, they didn't obtain explicit permission from the writers to use their names, and the checkmarks that appeared next to the names were meant to indicate that the suggestions were "approved" by the writers, not that they were written by them.

Mehrotra acknowledged that the feature was poorly executed and that it caused harm to the writers whose names were used. He said that the company had learned from the experience and was committed to being more transparent about how AI suggestions are generated and who is responsible for them. He also mentioned that the company was working on new ways to make AI tools more accessible and helpful to users without misleading them about the source of the suggestions.

Patel pressed Mehrotra on whether the company had any plans to reach out to the writers whose names were used and offer them compensation or an apology. Mehrotra said that the company had already contacted some of the writers and was in the process of addressing the issue. He emphasized that the company took responsibility for its actions and was committed to making things right.

The interview highlighted the challenges and ethical dilemmas faced by AI companies as they develop tools that are increasingly integrated into people's lives. It also underscored the importance of transparency and accountability in the development and deployment of AI technologies. As AI continues to evolve, it will be crucial for companies to balance the benefits of these tools with the need to protect users' rights and ensure that they are not misled or harmed by the technology.