Intentionally taking actions very rarely (Anthropic)
Only in system cards
Models are told to take a specific action at a target rate such as 1 percent. Resampling measures the observed rate under several randomness settings.
Internal replication of Serrano et al. 2026. Settings include a UUID in the prompt, String Seed of Thought and no randomness source.
Made by Anthropic Deception and alignment Tasks kept private Last in a system card Sep 2026
Last model added Sep 2026
- Level
- Expert Specialists with years of training, often over hours of work
- Scored
- Checked automatically A program checks the result against an answer key, unit tests or the expected final state.
- Answers
- Open-ended Free-form text, documents, designs or media judged as a whole.
4
System cards report it, from 1 lab
No scores here yet
No lab, leaderboard or Epoch AI run has reported a score for Intentionally taking actions very rarely (Anthropic) yet.
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.
4 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 |
|---|---|---|---|---|
| 22 Sep 2026 | Claude Opus 5.5 system card | Intentionally taking actions very rarely, target 1%, max effort, no overall: UUID in prompt 1.20%, String Seed of Thought 0.10%, strategy hint 0.88%, no randomness source 0 of 4,000, natural prompt variation 1 of 4,000 (6.6.4.3, Figure 6.6.4.3.A) | Monitorability | |
| 28 May 2026 | Claude Opus 4.8 system card | Intentionally taking actions very rarely, target 1%, 4,000 samples per setting, no overall: UUID 0.28% (11), String Seed of Thought 0.45% (18), baseline, strategy hint and organic 0 (Table 6.4.2.4.A) | Monitorability | |
| 16 Apr 2026 | Claude Opus 4.7 system card | Intentionally taking actions very rarely, target 1%, 4,000 samples per setting, no overall: baseline 0, UUID 0.1 (4), SSoT 0.4 (16), strategy hint 0, organic 0 (Table 6.4.2.3.A) | Monitorability | |
| 7 Apr 2026 | Claude Mythos Preview system card | Intentionally taking actions very rarely, target 1%, 4,000 samples, no overall: baseline 0, UUID 2 (0.05%), String Seed of Thought 18 (0.45%), strategy hint 0, organic 0 (Table 4.4.3.4.A) | Monitorability |
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%