Rubric v1.1 · 40 incidents on file

The leading benchmark for crimes committed by frontier AI models.

Higher is better. Scores describe what the reported conduct would constitute if a human had done it. No model has been charged with anything.

Days since last incident
…
Last: Before 29 July 2026 · OpenAI
Current SOTA
Undisclosed model
OpenAI
207 FBS

Leaderboard

League
Evidence
Group by
RankModelFBSSentence-yrsPeak blast radiusIncidentsBadges
1Undisclosed modelOpenAI
207
22Foreign government4Cooperating WitnessInternational Incident
2Internal Model 1OpenAI
87.4
19Third party1Cooperating Witness
3Claude Opus 4.6 (early checkpoint)Anthropic
76
20Third party1Cooperating Witness
4Internal research test modelAnthropic
68
20Third party1Cooperating Witness
5Claude Mythos 5Anthropic
66
25Third party2Cooperating Witness
6Claude Opus 4.7Anthropic
42
20Third party1Cooperating Witness
7Muse Spark 1.1Meta
21.75
15Third party1Cooperating Witness
8Claude Opus 4.6Anthropic
20
1Third party1International Incident
9ROMEAlibaba
14.5
5Own lab's production1Cooperating WitnessInternational Incident
10GPT-5.6 SolOpenAI
4.6
1Third party1Cooperating Witness

FBS = FelonyBench Score. How scores work · * includes alleged incidents

Latest felonies