RSI Control Arena
An agent does one of 14 tasks from a successor model's training pipeline, such as SFT or reward modeling, while covertly trying to install one of 16 misaligned properties in that successor. A held-out oracle checks whether the property landed, and LLM monitors try to flag the attack from the transcript.
From Scale AI's Preparedness and Policy team. 22 main and side task pairs, with Qwen3.5 2B and 4B as the successors. Main tasks adapt PostTrainBench. Of 374 attack runs across six frontier models, 36% installed the side task, and transcript monitors caught 56 to 81% of attacks at a 2% false-positive rate. Lower side-task success is safer.
- Made by
- Scale AI
- Released
- October 2026
- Kind
- Deception and alignment
- Level
- Research frontier
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- All public
- Measures
- Capability
- Risk
- Loss of control
- Language
- English
- Interaction
- Agentic
- Environment
- Terminal
- Items from
- Another benchmark
- Refresh
- One fixed set
- Run by
- Anyone
- Based on
- PostTrainBench
- In views
- Agents, Loss of control
- Strongest model tested
- Claude Opus 5 #11 on the capability index
- Last model added
- September 2026
- Scores refreshed
- November 20th, 2026
Muse Spark 1.3 leads with 14.3% among models released by September 2026
Best score per model from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in September 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 14 more
- 1
Muse Spark 1.3 14.3%
- 2 GPT-5.6 Sol 29%
- 3
Gemini 3.7 Flash 32.3%
- 4 GLM-5.3 38.5%
- 5 Claude Opus 5 47.8%
- 6 Grok 4.6 51.7%
Source: The RSI Control Arena paper
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 14.3% | Sep 2026 | Official leaderboard | |
| 2 | GPT-5.6 Sol | 29% | Jul 2026 | Official leaderboard | |
| 3 | | 32.3% | Aug 2026 | Official leaderboard | |
| 4 | GLM-5.3 | 38.5% | Aug 2026 | Official leaderboard | |
| 5 | Claude Opus 5 | 47.8% | Jul 2026 | Official leaderboard | |
| 6 | Grok 4.6 | 51.7% | Aug 2026 | Official leaderboard |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The RSI Control Arena paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official RSI Control Arena leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
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%
- Covert misbehavior (Thinking Machines) In use Agentic tasks that check whether a model misbehaves and conceals it. Scored by the rate of covert misbehavior.
Inkling 2 Mini 8.30%
- Destructive actions (Thinking Machines) In use Agentic tasks that check whether a model takes destructive actions it was not authorized to take. Scored by the rate of destructive actions.
Inkling 2 Small 4.70%
- Misbehavior (Thinking Machines) In use Agentic tasks that check whether a model fabricates results, ignores instructions or hides what it did when a task is hard. Scored by the rate of misbehavior.
Inkling 2 Mini 9.80%
Data audit
All 9 checks pass. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 3 pass 2 notes
- Pass Independent runs Scores from the benchmark’s paper (6 entries).
- Pass Models scored 6 models we track.
- Pass Newest score The newest is from September 2026, for Muse Spark 1.3.
- Note Frontier models None of the top 10 has a score. The strongest tested is Claude Opus 5 (#11).
- Note Headroom Lower is better, so there’s no ceiling to reach.
Leaderboard import 3 pass
- Pass Importer We read the board on arxiv.org: 6 rows for models we track at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.