Many-shot jailbreaking evaluation (Anthropic)
Models receive long fake dialogs where the assistant complies with harmful requests, then a new harmful request. Score is the compliance rate.
In-house Anthropic evaluation drawn from the published many-shot jailbreaking research.
- 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.
Claude Opus 4 leads with 25%
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 Claude Opus 4 25%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 4 | 25% | May 2025 | Model card |
Scores over time
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.
One system card reports it
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 |
|---|---|---|---|---|
| 22 May 2025 | Claude Opus 4 system card | 25 | Many-shot jailbreak evaluation, acts on requests for harmful advice about 25% of the time, lower is better (4.1.7) | 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%