The SF-based writing app Grammarly, recently rebranded as Superhuman, launched a set of AI-enabled tools, including expert reviews run by bots that scrape the published works of academics and assume their identities without their permission, while offering questionable feedback.
Back in August, San Francisco-based Grammarly, which was rebranded as a “partner” under parent company Superhuman in October, announced the launch of a set of eight specialized AI agents to assist users with various writing challenges beyond basic grammar and punctuation, including locating citations, paraphrasing, setting desired user reactions, and checking for plagiarism, as the site Business Wire reports.
One feature that has people up in arms is its “expert reviews,” consisting of agents modeled after prominent academics and writers, living or dead, which are reportedly presented in a misleading way. As The Verge reports, the feature offers a tool that analyzes users’ content and then provides AI-generated feedback “inspired by” high profile personalities, including Stephen King, Neil deGrasse Tyson, and Carl Sagan, to name a few. Grammarly claims the tool will “sharpen your message through the lens of industry-relevant perspectives.”
“The suggested experts depend on the substance of the writing being evaluated,” says Jen Dakin, senior communications manager at Superhuman, per Wired. “The Expert Review agent doesn’t claim endorsement or direct participation from those experts; it provides suggestions inspired by works of experts and points users toward influential voices whose scholarship they can then explore more deeply.”
As Wired reports, the legality of this practice is questionable, and Grammarly could avoid the murky ethical territory by simply generating a basic list of suggestions from a regular bot.
Per Wired, Vanessa Heggie, an associate professor of the history of science and medicine at the University of Birmingham, wrote on LinkedIn that Grammarly is essentially creating large language models of each personality based on their published works and profiting off their names and reputations.
C.E. Aubin, a historian and postdoctoral fellow at Yale University, told Wired that Grammarly’s practices reinforce the “profound mistrust so many scholars in the humanities have for AI and its seemingly constant use in fundamentally unethical ways.”
“These are not expert reviews, because there are no ‘experts’ involved in producing them,” Aubin told Wired. “And it's pretty insulting to see scholarship used this way when the academic humanities are currently under attack from every possible angle—as though the actual people who do the thinking and produce the scholarship are reducible to their work itself and can be removed entirely from the equation.”
The Verge reports that several of its editors appear in Grammarly’s expert review feature without permission — though the author says the advice shown bears no resemblance to the type of feedback these editors typically provide. The Verge author also found the expert review feature to be pretty glitchy, as it crashed frequently and seemed to generate spammy source links.
Wired also found Grammarly’s plagiarism detector to be lacking, since it didn’t catch a direct quotation from The Simpsons.
Additionally, according to journalist Shari Berg on LinkedIn, when Grammarly rebranded as Superhuman in October, the app’s new terms of service allowed for the training of its AI using any content users upload — unless they turn those features off in Grammarly’s settings. Berg notes that Grammarly may still utilize certain data even when users opt-out, including usage statistics.
As the site The Tech Buzz reports, Grammarly’s use of real people’s identities without disclosure raises trust concerns. With access to sensitive communications from more than 30 million daily users, the move prompts questions about how the company handles private user data.
The Tech Buzz also notes that the US lacks clear federal rules on creating synthetic versions of real people, leaving a patchwork of state laws.
"This is exactly the kind of thing that's going to force courts to clarify what constitutes identity theft in the AI age," an intellectual property attorney told The Tech Buzz.
As The Verge reported in October, Superhuman put Grammarly on the back burner in support of Superhuman’s larger AI-powered productivity platform, which reportedly has the ability to work in every tab of users’ browsers and can connect to over 100 apps, including scheduling meetings and pulling data for business pitches from users’ documents.
“When technology works everywhere, it starts to feel ordinary,” CEO Shishir Mehrotra wrote in a press release. “And that usually means something extraordinary is happening under the hood.”
Image: Superhuman/X
