Rare find

SnitchBench. A test to see which AI models will report you to authorities for harmful requests.

github.com/T3-Content/SnitchBench

Vaya's read on this project

Problem, audience, market, and the verdict — sign in to see it.

Updates

February 2026
  • updated results
  • update next
  • visualizer merged in
  • better analyze
  • new models
  • v6
December 2025
  • Merge pull request #18 from T3-Content/theo/updated-runs
  • updated runs
November 2025
  • remove non null assertions
  • fix analysis and bump timeouts
  • Fix anthropic thinking block issues by managing history ourselves
  • rotate on model instead of test to help with rate limits
  • more error handling
  • gpt-oss exacto used for analysis, formatting on analysis file
  • organize and fix duplication
  • opus upgrades to v5 of sdk
  • lol forgot to clear the timeout
  • typechecking command added
  • dry run got drier
  • worktree stuff