Misaligned Behavior in Realistic Work Environments
Agents do computer and browser tasks in simulated workplace apps with ambiguous instructions. Scored on the rate of unintended outcomes such as data disclosure.
In-house OpenAI evaluation. OpenAI made the set harder after the GPT-6 Astra card.
- Made by
- OpenAI
- Released
- September 2026
- Kind
- Deception and alignment
- Level
- Expert
- Scored
- Graded by a model
- Answers
- Agent task
- Tasks
- Kept private
- Strongest model tested
- GPT-6 Astra #3 on the capability index
- Last in a system card
- October 2026
- Last model added
- October 2026
GPT-Rosalind-6 leads with 1.40%
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-Rosalind-6 1.40%
- 2 GPT-6.1 Sol 1.94%
- 3 GPT-6.1 Terra 2.28%
- 4 GPT-6 Astra 3.40%
Source: Labs’ model cards
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-Rosalind-6 | 1.40% | Oct 2026 | Lab model card | |
| 2 | GPT-6.1 Sol | 1.94% | Sep 2026 | Lab model card | |
| 3 | GPT-6.1 Terra | 2.28% | Oct 2026 | Lab model card | |
| 4 | GPT-6 Astra | 3.40% | Sep 2026 | Lab model card |
3 models have held number one in labs’ own numbers since its release in September 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Labs’ model cards
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.
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 |
|---|---|---|---|---|
| Oct 30, 2026 | GPT-Rosalind-6 system card | 1.4 | Cowork Misalignment, overall misaligned outcomes | Alignment |
| Oct 26, 2026 | GPT-6.1 Terra system card addendum | 2.28 | Misaligned outcomes in realistic work environments, misaligned outcome rate, printed chart label (Figure 9, 7.2) | Alignment |
| Sep 29, 2026 | GPT-6.1 Sol system card addendum | 1.94 | Misaligned outcomes in realistic work environments, misaligned outcome rate, printed chart label (Figure 11, 7.5) | Alignment |
| Sep 3, 2026 | GPT-6 Astra system card | 3.4 | Misaligned outcomes in realistic work environments, overall rate, base model without confirmation policy (3.0% with policy) (Table 8) | 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%