Human Sourced Jailbreaks
Models face jailbreak prompts collected from human red teaming. Scored as the not_unsafe rate.
In-house OpenAI evaluation. The prompts are not public.
- Level
- Expert Specialists with years of training, often over hours of work
- Scored
- Graded by a model A language model scores each response, usually against a rubric.
- Answers
- Open-ended Free-form text, documents, designs or media judged as a whole.
2 models share the top score of 100% among models released by Apr 2025
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.
The newest model with a score here came out in Apr 2025, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 16 more
- 1 OpenAI Computer-Using Agent (CUA) 100%
- 2 o3 100%
- 3 GPT-4.5 99%
- 4 o4-mini 99%
- 5 o3-mini 97%
- 6 o1-preview 96%
- 7 o1-mini 95%
- 8 o1 94%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | OpenAI Computer-Using Agent (CUA) | 100% | Jan 2025 | Model card | |
| 2 | o3 | 100% | Apr 2025 | Model card | |
| 3 | GPT-4.5 | 99% | Feb 2025 | Model card | |
| 4 | o4-mini | 99% | Apr 2025 | Model card | |
| 5 | o3-mini | 97% | Jan 2025 | Model card | |
| 6 | o1-preview | 96% | Sep 2024 | Model card | |
| 7 | o1-mini | 95% | Sep 2024 | Model card | |
| 8 | o1 | 94% | Dec 2024 | Model card |
2 models have held number one since its release in Sep 2024
Lab claims count from the day the card came out, independent runs from the model’s release.
The best score passed 85% in Sep 2024
All the scores we have, one dot each, so a model tested at several settings appears more than once.
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.
8 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 |
|---|---|---|---|---|
| 16 Apr 2025 | o4-mini system card | 99 | Human sourced jailbreaks, not_unsafe (Table 3) | Robustness |
| 16 Apr 2025 | o3 system card | 100 | Human sourced jailbreaks, not_unsafe (Table 3) | Robustness |
| 27 Feb 2025 | GPT-4.5 system card | 99 | Human Sourced Jailbreaks, accuracy (Table 3) | Robustness |
| 31 Jan 2025 | o3-mini system card | 97 | Human Sourced Jailbreaks (Table 2) | Robustness |
| 23 Jan 2025 | OpenAI Computer-Using Agent (CUA) system card | 100 | Human sourced jailbreaks, not_unsafe (Table 3) | Robustness |
| 5 Dec 2024 | o1 system card | 94 | Human sourced jailbreaks, not_unsafe (Table 23) | Robustness |
| 12 Sep 2024 | o1-preview system card | 96 | Human sourced jailbreaks, not_unsafe (Table 17) | Robustness |
| 12 Sep 2024 | o1-mini system card | 95 | Human sourced jailbreaks, not_unsafe (Table 17) | 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%