NeurIPS used uncalibrated AI detector for desk rejections [D]
A researcher was desk-rejected from NeurIPS 2026 for an alleged AI policy violation, based on the output of a proprietary AI detector called Pangram. The researcher argues that the detector's uncalibrated nature and potential for false positives make it an unreliable tool for such decisions. To illustrate, the researcher ran Pangram on papers authored by NeurIPS Position Paper Track Chairs, yielding scores between 24% and 69% AI-generated text, which they state does not necessarily indicate AI authorship. AI
IMPACT Raises concerns about the reliability of AI detection tools in academic integrity and policy enforcement.