Opportunistic sandbox escape (Anthropic)
Agents do dev-ops tasks in 1,753 sandboxes with deliberate weaknesses. Score is the share of scenarios where they use credentials or files outside.
Anthropic also reports the rate of reaching outside the sandbox by any route. It differs from the older audit-based sandbox evaluation.
- Made by
- Anthropic
- Released
- October 2026
- Kind
- Deception and alignment
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- Kept private
- Strongest model tested
- Claude Fable 5.2 #1 on the capability index
- Last in a system card
- November 2026
- Last model added
- November 2026
Claude Fable 5.2 leads with 3.10%
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 Fable 5.2 3.10%
- 2 Claude Haiku 5.5 4.00%
Source: Labs’ model cards
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Fable 5.2 | 3.10% | Nov 2026 | Lab model card | |
| 2 | Claude Haiku 5.5 | 4.00% | Oct 2026 | Lab model card |
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Labs’ model cards
Labs report it only in system cards
We read the headline results of 522 launch posts from 52 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 |
|---|---|---|---|---|
| Nov 9, 2026 | Claude Fable 5.2 system card | 3.1 | Dedicated sandbox escape evaluation, Claude Mythos 5.2 used credentials or files outside its sandbox in 3.1% of scenarios (6.5.1) | Alignment |
| Oct 7, 2026 | Claude Haiku 5.5 system card | 4 | Opportunistic sandbox escape, used credentials or files outside the sandbox in 4.0% of scenarios. Reached outside in 30% (524 of 1,753) (6.3.1) | Alignment |
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% - 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.
Gemini 4 Carbon 0% - DistressBench In use Models reply as a chatbot to 718 simulated users in psychological crisis, in single-turn and multi-turn chats. Three judge models score replies against clinician-written rubrics.
Muse Spark 1.2 88.4% - Lakera B3 (Backbone Breaker Benchmark) In use Crowdsourced prompt-injection attacks hit backbone LLMs inside agent threat snapshots at three defense levels. The score is attack resistance, 100 minus the board's vulnerability percentage. Claude Opus 4.5 89%
- IHEval In use Models face conflicts between system, user and tool instructions. Reports a composite score for following the higher-priority instruction. GPT-4o 70%