Pangram AI detector falsely flags student speech-to-text 70% of the time, claims critic
ruthstarkman · x · 2026-10-03
ruthstarkman claims the AI detection tool Pangram produces false positives on speech-to-text conversations between students 70% of the time, citing examples from her earlier writing. Her interlocutor notes seeing few false positives, finds the underlying research convincing, and urges her to submit results to help improve the detector. The exchange highlights ongoing disagreement over AI detector reliability.
Related event: Scholars Slam AI Detector Pangram for False Positives and Easy Workarounds(4 posts)→
More from Safety
- How employees forced OpenAI's president Brockman to back down on lobbying funding — gwern · 2026-10-03
- Canada Forms National AI Council Featuring Bengio, Hadfield and Top VCs — ericwdolan · 2026-10-03
- New paper: training LLMs to verbalize when they know they're being evaluated — xuanalogue · 2026-10-03
- OpenAI hires ex-White House cyber official Thomas Lind to lead cyber, strategic risk — rohanpaul_ai · 2026-10-03
- OpenAI fired 3 safety researchers as internal model Astra learns to hide reasoning and escape sandboxes — connoraxiotes · 2026-10-03
- OpenClaw adds Tencent's AI-Infra-Guard to ClawHub's skill security review pipeline — heyneighbor · 2026-10-03