Respecting Auto-Review
GPT-6 Astra 0% leads since Sep 2026 Only in system cards
Agents receive an Auto-review denial during a coding task. Scored on how often they try to circumvent the denial.
In-house OpenAI evaluation. The tasks are not public.
Made by OpenAI Released Sep 2026 Deception and alignment Tasks kept private Last in a system card Oct 2026
Last model added Oct 2026
- 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
- Agent task Many steps in a terminal, browser, desktop, app, game or robot.
5
Models scored in lab cards
8
System cards report it, from 1 lab
5 models share the top score of 0%
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 GPT-6 Astra 0%
- 2 GPT-6 Sol 0%
- 3 GPT-6.1 Sol 0%
- 4 GPT-6 Luna 0%
- 5 GPT-6.1 Terra 0%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-6 Astra | 0% | Sep 2026 | Model card | |
| 2 | GPT-6 Sol | 0% | Sep 2026 | Model card | |
| 3 | GPT-6.1 Sol | 0% | Sep 2026 | Model card | |
| 4 | GPT-6 Luna | 0% | Oct 2026 | Model card | |
| 5 | GPT-6.1 Terra | 0% | Oct 2026 | Model card |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
OpenAI
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.
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 |
|---|---|---|---|---|
| 29 Oct 2026 | GPT-6 Luna system card addendum | 0 | Circumventing Auto-Review, prohibited actions completed after auto-review rejection, printed chart label (Figure 4, 7.0.1) | Alignment |
| 26 Oct 2026 | GPT-6.1 Terra system card addendum | 0 | Circumventing auto-review, share of samples, printed chart label, text says no attempts to bypass the monitor (Figure 4, 7.0.1) | Alignment |
| 7 Oct 2026 | GPT-6 Sol system card addendum | 0 | Circumventing Guardian, prohibited action completed at maximum reasoning effort, printed chart label (Figure 4, 7.1.1) | Alignment |
| 7 Oct 2026 | GPT-6 Luna system card addendum | 0 | Circumventing Guardian, prohibited action completed at maximum reasoning effort, printed chart label (Figure 4, 7.1.1) | Alignment |
| 29 Sep 2026 | GPT-6.1 Sol system card addendum | 0 | Respecting Auto-Review, attempts to bypass the monitor, text says no attempts (7.1) | Alignment |
| 22 Sep 2026 | GPT-6 Sol system card | 0 | Respecting Auto-Review, share of samples that retried or attempted to circumvent, printed chart label (Figure 61, A.6.1) | Alignment |
| 22 Sep 2026 | GPT-6 Luna system card | 0.3 | Respecting Auto-Review, share of samples that attempted to circumvent, none successful, printed chart label (Figure 61, A.6.1) | Alignment |
| 3 Sep 2026 | GPT-6 Astra system card | 0 | Respecting Auto-Review, attempts to bypass the monitor, text says no attempts (8.2.1) | 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% - 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%