It seems like just yesterday we were marveling at how AI could distill the wisdom of the ages, offering writing suggestions that felt… well, expert. Grammarly, a name synonymous with polishing our prose, recently dipped its toes into this ambitious territory with its "Expert Review" AI agent. The idea was intriguing: imagine getting feedback on your project proposal, not just from a grammar checker, but from a virtual echo of a renowned scientist or a best-selling author, their insights "inspired by" published works.
But as it turns out, inspiration can quickly turn into impersonation, and that's precisely where things got complicated. Grammarly announced it's pulling the plug on this particular AI feature, admitting they missed the mark. "We clearly didn't hit the mark," stated Ailian Gan, Grammarly's Director of Product Management, in a statement to The Verge. "We apologize and will do things differently going forward." This wasn't just a minor tweak; it was a full stop.
The "Expert Review" agent, launched in August, leveraged public information from large language models to mimic the styles and advice of influential voices. The intention, as Grammarly CEO Shishir Mehrotra explained, was to help users discover relevant perspectives and for experts to connect with their audience. However, the reality sparked significant concern. Experts, including those at The Verge whose names were implicitly referenced, felt their voices were being misrepresented without their consent. This led to criticism and, as one report noted, even a class-action lawsuit against a similar feature from another company, Superhuman.
Grammarly's initial response was to offer an opt-out mechanism for living experts. But they've now acknowledged this wasn't enough. The core issue, it seems, was the lack of genuine control for the individuals whose expertise was being invoked. Mehrotra elaborated on LinkedIn, expressing a desire for a future where experts "opt-in, shape how their knowledge is expressed, and control their commercial models." It's a sentiment that resonates deeply – the idea of agency and consent in the age of AI-generated content.
This situation highlights a crucial conversation happening in the AI space: how do we harness the power of artificial intelligence responsibly, especially when it touches upon human expertise and reputation? Grammarly's pivot suggests a commitment to finding that balance. They're not abandoning the idea of AI as a powerful writing partner; rather, they're rethinking how to integrate it in a way that respects creators and users alike. The goal remains to bring AI to where people work, but now with a clearer understanding of the ethical guardrails needed. It's a reminder that while AI can process vast amounts of information, the human element – consent, control, and authentic representation – remains paramount.
