Two paper authors complained on Nov. 6, 2026, that Pangram had flagged their peer reviews as AI-generated. Their posts describe reported detector results and their frustration with the reviews, rather than established findings about how the reviews were written.
In one post, @nsaphra wrote that a paper they were excited about received scores of “2/2/3/2” and that “Pangram flagged 3/4 as 100% AI generated.” The post did not name a conference or explain the scoring scale.
“I hate that we likely can't afford to ban these reviewers from refereeing,” @nsaphra wrote. That was the author's view, not an announcement of sanctions.
In a separate post, @TuhinChakr explicitly identified the reviews as ICLR reviews, reporting Pangram ratings of 91%, 93% and 91% AI-generated. The author also alleged that one review was “a simple Claude Copy paste with the most elementary nitpick.” That characterization does not establish that Claude produced the review.
“It’s insulting and dehumanizing for authors who do honest work,” @TuhinChakr wrote, adding, “I think as a community we are underestimating what GenAI is doing to our conferences.”
The percentages remain the authors' accounts of Pangram's outputs, not verified findings of AI authorship. Neither post includes the review text or detector reports.
