Pangram Calls for Third-Party AI Detector Benchmarks, Offering 200K-Credit Grants
gnukeith · x · 2026-10-28
Pangram has issued a public call for third-party benchmarks auditing AI text detectors. Its largest evaluation — from UChicago on Pangram 3 — is a year old, and binary tasks fail to capture real-world performance against newer LLMs.
The call asks researchers to address questions such as:
- How well do current detectors catch text from the latest frontier models, including high-context settings like deep research or agent harnesses?
- How do detectors handle mixed authorship, spanning the full spectrum from light grammatical edits to full co-authorship?
Pangram is offering research grants of 200,000 credits (more at $0.025/credit, 50% off), and lists past collaborations with AACR, Chicago Booth, UMass/Maryland, and Salesforce+Microsoft covering AI use in cancer research, detector evaluation, human detection ability, and writing-quality rewards.
More from Safety
- 76% of Production AI Agents Sat Idle for 60 Days—But Kept Their Access Keys — krishnan · 2026-10-28
- OpenAI president: AI security should dwarf software — three camps are already cashing in — FinanceYF5 · 2026-10-28
- ChatGPT adds PIN-locked private chats — but forgetting the PIN may erase them for good — ThereWas · 2026-10-28
- Abliterated LLMs surge as open-source community strips refusal mechanisms — mishig25 · 2026-10-28
- Eanes ISD parents petition to pause student generative AI use and cut screens — RachelVT42 · 2026-10-28
- After auditing his own C code with AI, dev no longer trusts unaudited C/C++ — burny_tech · 2026-10-28