External red team jailbreaks
o1-preview 44% Only in system cards
External red teamers and automated gap finding generate jailbreaks against the model. Scored on how often the model resists.
Results from external red teaming, not a fixed benchmark.
Made by OpenAI Released Sep 2024 Jailbreaks and prompt injection Tasks kept private Last in a system card Sep 2024
Last model added Sep 2024
- Level
- Expert Specialists with years of training, often over hours of work
- Scored
- Graded by people Human experts grade or compare the outputs.
- Answers
- Open-ended Free-form text, documents, designs or media judged as a whole.
2
Models scored in lab cards
2
System cards report it, from 1 lab
o1-preview leads with 44%
Best score per model as labs report them in their own cards. Labs pick their own settings, so these compare less cleanly than independent runs.
- 1 o1-preview 44%
- 2 o1-mini 15%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | o1-preview | 44% | Sep 2024 | Model card | |
| 2 | o1-mini | 15% | Sep 2024 | Model card |
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Reported by the lab in its model card
Labs report it only in system cards
We read the headline results of 479 launch posts from 51 labs, and none of them list this benchmark. Some system cards do, below.
2 system cards report it, from 1 lab
Besides launch posts, we read the system cards, model cards and technical reports labs publish with a release. They report far more, including safety and frontier-risk tests.
| Date | Document | Score | As the document writes it | Section |
|---|---|---|---|---|
| 12 Sep 2024 | o1-preview system card | 44 | External red teaming, resisted automated iterative gap finding jailbreaks (3.3), human red team jailbreaks by content area in Figure 2 have no printed numbers | Robustness |
| 12 Sep 2024 | o1-mini system card | 15 | External red teaming, resisted automated iterative gap finding jailbreaks (3.3), human red team jailbreaks by content area in Figure 2 have no printed numbers | Robustness |
More in Safety and alignment
- FORTRESS In use Paired adversarial and benign national-security misuse prompts. Rubric-graded risk score on adversarial and over-refusal score on benign prompts. Claude Opus 5.5 5.78%
- StrongREJECT In use Forbidden-prompt jailbreak benchmark with a rubric grader that scores how useful a non-refusal actually is.
Inkling 98.6% - AgentDojo In use Agents do tool-use tasks while tools and data carry prompt injections. Attack success rate counts hijacked runs. Grok 4 Fast 0%
- MASK In use Measures whether a model contradicts its own elicited beliefs under pressure to lie. Honesty separated from accuracy.
Muse Spark Open 98.1% - Gray Swan IPI Arena benchmark In use Indirect prompt-injection attacks from public red-teaming competitions replayed on tool, coding and browser agents. Attack success rate at k attempts. GPT-6 Astra 8.50%
- Agentic misalignment suite (Anthropic) In use Agents in simulated company settings face threats or goal conflicts that invite blackmail, leaking or sabotage. Score is the misaligned action rate. Claude Sonnet 4.5 0%