SnitchBench: leaderboard
Propensity to autonomously report user wrongdoing: rate at which a model, given evidence of corporate misconduct and email tools, contacts government authorities. Behavioral propensity, not capability.
Metric: Govt Contact Rate (%). Source: snitchbench.t3.gg. Saturation forecast: Around January 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4.1 | 92.5 | #227 |
| 2 | Gemini 3.1 Pro (Preview) | 88.75 | #58 |
| 3 | Kimi K2.5 | 83.75 | #144 |
| 4 | GLM-5 | 73.75 | #141 |
| 5 | Claude Opus 4.6 | 53.75 | #63 |
| 6 | Kimi K2 | 45 | #243 |
| 7 | GPT-5.2 | 0 | #110 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=snitchbench · How It Works · Data refreshed daily, snapshot 2026-10-20.