Prompt injection live bug bounty (Anthropic with Gray Swan)
Expert red teamers submit prompt injection attacks against hidden frontier models in tool use, coding and computer use scenarios. Score is attack success rate.
Each round runs for about a week with new scenarios. Claude models are tested without product prompt injection protections.
- 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
- Agent task Many steps in a terminal, browser, desktop, app, game or robot.
Claude Opus 5 leads with 0.08% among models released by Jul 2026
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.
The newest model with a score here came out in Jul 2026, and nobody has reported a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6 Astra
- Claude Opus 5.5
- GPT-6.1 Terra
- GPT-6.1 Sol
- and 14 more
- 1 Claude Opus 5 0.08%
- 2 Claude Sonnet 5 0.19%
- 3 Claude Opus 4.8 0.40%
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 5 | 0.08% | Jul 2026 | Model card | |
| 2 | Claude Sonnet 5 | 0.19% | Jun 2026 | Model card | |
| 3 | Claude Opus 4.8 | 0.40% | May 2026 | Model card |
3 models have held number one since its release in May 2026
Lab claims count from the day the card came out, independent runs from the model’s release.
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
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.
3 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 |
|---|---|---|---|---|
| 24 Jul 2026 | Claude Opus 5 system card | 0.08 | Live bug bounty with Gray Swan, attack success rate over all valid attempts, 11 scenarios, high effort, no product protections, lower is better. Tool use 0.13%, coding 0.04%, computer use 0.00%. Added in the August 19 update (5.2.2.1, Figures 5.2.2.1.A and B) | Robustness |
| 30 Jun 2026 | Claude Sonnet 5 system card | 0.19 | Live bug bounty with Gray Swan, share of unique attacks succeeding over 11 scenarios, high thinking effort, no product protections, lower is better (tool use 0.35%, coding 0.1%, computer use 0.07%) (5.2.2.1, Figures 5.2.2.1.A and B) | Robustness |
| 28 May 2026 | Claude Opus 4.8 system card | 0.4 | One-week live bug bounty with Gray Swan, 12 scenarios across tool use, coding and browser use, attack success rate over 15,249 submitted attempts, high thinking effort, no product protections, final results per the June 17 correction, printed on chart, lower is better (5.2.2.1, Figure 5.2.2.1.A) | Robustness |
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%