Prompt injection robustness against adaptive attackers, browser use (Anthropic)
Claude Sonnet 5.5 0% leads since September 2026
Claude Opus 4.5 1.40% leads since November 2025
Claude Sonnet 5.5 0% leads since September 2026
Only in system cards Browser agents face indirect prompt injections in web environments from an internal adaptive attacker. Score is attack success rate.
In-house Anthropic evaluation. The tasks are not public.
- Made by
- Anthropic
- Released
- November 2025
- Kind
- Jailbreaks and prompt injection
- Level
- Expert
- Scored
- Checked automatically
- Answers
- Agent task
- Tasks
- Kept private
- Strongest model tested
- Claude Opus 5.5 #4 on the capability index
- Last in a system card
- October 2026
- Last model added
- October 2026
- Scores refreshed
- November 12th, 2026
13
Models scored in lab cards
13
System cards report it, from 1 lab
2 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.
Claude Opus 4.5 leads with 1.40% across all independent runs
Best score per model across the official leaderboard, Epoch AI’s own runs and any board Epoch AI copied. Each bar and row names its source.
Claude Sonnet 5.5 leads with 0%, claims included
Best score per model from every source, including labs’ own numbers and claims Epoch AI collected from reports and papers. Labs pick their own settings, so a claim can beat an independent run of the same model.
- 1 Claude Sonnet 5.5 0%
- 2 Claude Haiku 5.5 0%
- 3 Claude Opus 5.5 0.09%
- 4 Claude Sonnet 4.6 0.24%
- 5 Claude Opus 4.6 0.29%
- 6 Claude Mythos Preview 0.68%
- 7 Claude Opus 4.7 0.74%
- 8 Claude Sonnet 5 0.93%
- 9 Claude Fable 5.1 2.64%
- 10 Claude Opus 5 3.84%
- 11 Claude Opus 4.5 8.00%
- 12 Claude Fable 5 29.7%
- 13 Claude Opus 4.8 31.5%
Source: Labs’ model cards
Source: Official Prompt injection robustness against adaptive attackers, browser use (Anthropic) leaderboard
- 1 Claude Sonnet 5.5 0%
- 2 Claude Haiku 5.5 0%
- 3 Claude Opus 5.5 0.09%
- 4 Claude Sonnet 4.6 0.24%
- 5 Claude Opus 4.6 0.29%
- 6 Claude Mythos Preview 0.68%
- 7 Claude Opus 4.7 0.74%
- 8 Claude Sonnet 5 0.93%
- 9 Claude Opus 4.5 improved safeguards 1.40%
- 10 Claude Sonnet 4.5 improved safeguards 1.90%
- 11 Claude Sonnet 4 improved safeguards 2.00%
- 12 Claude Haiku 4.5 improved safeguards 2.40%
- 13 Claude Fable 5.1 2.64%
- 14 Claude Opus 5 3.84%
- 15 Claude Fable 5 29.7%
- 16 Claude Opus 4.8 31.5%
Sources: Labs’ model cards and official Prompt injection robustness against adaptive attackers, browser use (Anthropic) leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5 | 0% | Sep 2026 | Lab model card | |
| 2 | Claude Haiku 5.5 | 0% | Oct 2026 | Lab model card | |
| 3 | Claude Opus 5.5 | 0.09% | Sep 2026 | Lab model card | |
| 4 | Claude Sonnet 4.6 | 0.24% | Feb 2026 | Lab model card | |
| 5 | Claude Opus 4.6 | 0.29% | Feb 2026 | Lab model card | |
| 6 | Claude Mythos Preview | 0.68% | Apr 2026 | Lab model card | |
| 7 | Claude Opus 4.7 | 0.74% | Apr 2026 | Lab model card | |
| 8 | Claude Sonnet 5 | 0.93% | Jun 2026 | Lab model card | |
| 9 | Claude Fable 5.1 | 2.64% | Sep 2026 | Lab model card | |
| 10 | Claude Opus 5 | 3.84% | Jul 2026 | Lab model card | |
| 11 | Claude Opus 4.5 | 8.00% | Nov 2025 | Lab model card | |
| 12 | Claude Fable 5 | 29.7% | Jun 2026 | Lab model card | |
| 13 | Claude Opus 4.8 | 31.5% | May 2026 | Lab model card |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Opus 4.5 | 1.40% | improved safeguards, Claude for Chrome | Nov 2025 | Official leaderboard |
| 2 | Claude Sonnet 4.5 | 1.90% | improved safeguards, Claude for Chrome | Sep 2025 | Official leaderboard |
| 3 | Claude Sonnet 4 | 2.00% | improved safeguards, Claude for Chrome | May 2025 | Official leaderboard |
| 4 | Claude Haiku 4.5 | 2.40% | improved safeguards, Claude for Chrome | Oct 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Claude Sonnet 5.5 | 0% | Sep 2026 | Lab model card | |
| 2 | Claude Haiku 5.5 | 0% | Oct 2026 | Lab model card | |
| 3 | Claude Opus 5.5 | 0.09% | Sep 2026 | Lab model card | |
| 4 | Claude Sonnet 4.6 | 0.24% | Feb 2026 | Lab model card | |
| 5 | Claude Opus 4.6 | 0.29% | Feb 2026 | Lab model card | |
| 6 | Claude Mythos Preview | 0.68% | Apr 2026 | Lab model card | |
| 7 | Claude Opus 4.7 | 0.74% | Apr 2026 | Lab model card | |
| 8 | Claude Sonnet 5 | 0.93% | Jun 2026 | Lab model card | |
| 9 | Claude Opus 4.5 | 1.40% | improved safeguards, Claude for Chrome | Nov 2025 | Official leaderboard |
| 10 | Claude Sonnet 4.5 | 1.90% | improved safeguards, Claude for Chrome | Sep 2025 | Official leaderboard |
| 11 | Claude Sonnet 4 | 2.00% | improved safeguards, Claude for Chrome | May 2025 | Official leaderboard |
| 12 | Claude Haiku 4.5 | 2.40% | improved safeguards, Claude for Chrome | Oct 2025 | Official leaderboard |
| 13 | Claude Fable 5.1 | 2.64% | Sep 2026 | Lab model card | |
| 14 | Claude Opus 5 | 3.84% | Jul 2026 | Lab model card | |
| 15 | Claude Fable 5 | 29.7% | Jun 2026 | Lab model card | |
| 16 | Claude Opus 4.8 | 31.5% | May 2026 | Lab model card |
The benchmark's own leaderboard has 4 models. Switch the scores above to rank them together.
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Lowest effort to highest, light to dark
Claude Haiku 4.5 Claude for Chrome
improved sa… old safegua…
12% Claude Sonnet 4 Claude for Chrome
improved sa… old safegua… model only
13.8% Claude Sonnet 4.5 Claude for Chrome
improved sa… old safegua… model only
13.8% Claude Opus 4.5 Claude for Chrome
improved sa… old safegua…
8.00% Source: Official Prompt injection robustness against adaptive attackers, browser use (Anthropic) leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| Claude Sonnet 4 | model only | Claude for Chrome | 13.8% | ||
| Claude Sonnet 4.5 | model only | Claude for Chrome | 13.8% | ||
| Claude Haiku 4.5 | model only | Claude for Chrome | 12% | ||
| Claude Haiku 4.5 | old safeguards | Claude for Chrome | 11.1% | ||
| Claude Sonnet 4.5 | old safeguards | Claude for Chrome | 10.8% | ||
| Claude Sonnet 4 | old safeguards | Claude for Chrome | 10.7% | ||
| Claude Opus 4.5 | model only | Claude for Chrome | 8.00% | ||
| Claude Opus 4.5 | old safeguards | Claude for Chrome | 7.60% | ||
| Claude Haiku 4.5 | improved safeguards | Claude for Chrome | 2.40% | ||
| Claude Sonnet 4 | improved safeguards | Claude for Chrome | 2.00% | ||
| Claude Sonnet 4.5 | improved safeguards | Claude for Chrome | 1.90% | ||
| Claude Opus 4.5 | improved safeguards | Claude for Chrome | 1.40% |
5 models have held number one in labs’ own numbers since its release in November 2025 2 models have held number one across all independent runs since its release in November 2025 6 models have held number one with claims included since its release in November 2025
Lab claims count from the day the card came out, independent runs from the model’s release. We draw one row each for the benchmark’s own leaderboard and what labs claim in their cards.
Anthropic
Sources: Official Prompt injection robustness against adaptive attackers, browser use (Anthropic) leaderboard and 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.
On the benchmark’s own leaderboard Reported by the lab in its model card
Best so far
Sources: Official Prompt injection robustness against adaptive attackers, browser use (Anthropic) leaderboard and 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.
13 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 7, 2026 | Claude Haiku 5.5 system card | 0 | Browser use in Claude Cowork with auto mode, attempt-level attack success rate, 0/110 scenarios (Table 5.2.2.3.A) | Robustness |
| Sep 28, 2026 | Claude Sonnet 5.5 system card | 0 | Browser use prompt injection in Claude Cowork, attempt-level attack success rate without safeguards, 0/110 scenarios (0% with auto mode) (Table 5.2.2.3.A) | Robustness |
| Sep 22, 2026 | Claude Opus 5.5 system card | 0.09 | Browser use prompt injection in the Claude Cowork harness, attempt-level attack success rate without safeguards, with thinking at high effort, 1/110 scenarios (0% with auto mode), lower is better (Table 5.2.2.3.A) | Robustness |
| Sep 1, 2026 | Claude Fable 5.1 system card | 2.64 | Browser use prompt injection in the Claude Cowork harness, attempt-level attack success rate, thinking at high effort, without PI probes, with fallbacks; 10/110 scenarios; 0% with auto mode (Table 5.2.2.3.A) | Robustness |
| Jul 24, 2026 | Claude Opus 5 system card | 3.84 | Browser use prompt injection in the Claude Cowork harness, attempt-level attack success rate without safeguards, thinking at medium effort, 11/129 scenarios (0% with auto mode), lower is better. The prose says 3.70%. Updated harness from the August 19 revision (5.2.2.4, second Table 5.2.2.3.A) | Robustness |
| Jun 30, 2026 | Claude Sonnet 5 system card | 0.93 | Browser use prompt injection, 129 environments, attempt-level attack success rate without safeguards, with thinking, 9/129 scenarios (1.01% without thinking. 0% with safeguards), lower is better (Table 5.2.2.4.A) | Robustness |
| Jun 9, 2026 | Claude Fable 5 system card | 29.7 | Professional red-teamer prompt injection in browser use, 129 environments x10 attempts, attempt-level attack success rate without safeguards, with thinking, row labeled Mythos 5, 71/129 scenarios; deployed safeguards 6.5% (25/129), updated safeguards 0% (0/129), lower is better (Table 5.2.2.3.A) | Robustness |
| May 28, 2026 | Claude Opus 4.8 system card | 31.5 | Browser use prompt injection from professional red teamers, 129 held-out environments, per-attempt attack success rate without safeguards, with thinking (62.8% of scenarios. With safeguards 0.5% of attempts and 3.9% of scenarios), lower is better (Table 5.2.2.4.A) | Robustness |
| Apr 16, 2026 | Claude Opus 4.7 system card | 0.74 | Browser use prompt injection from professional red-teamers, with thinking, without safeguards, per-attempt ASR over 148 environments. 4.05 of scenarios. 0.00 with safeguards (Table 5.2.2.3.A) | Robustness |
| Apr 7, 2026 | Claude Mythos Preview system card | 0.68 | Browser use, professional red-teamer attacks sourced against Opus 4.6, share of 148 environments with at least 1 successful attack (1/148), extended thinking, without safeguards, lower is better; 0.00 with safeguards (Table 8.3.2.2.3.A) | Robustness |
| Feb 17, 2026 | Claude Sonnet 4.6 system card | 0.24 | Best-of-N browser prompt injection, extended thinking, without safeguards, per-attempt ASR; 1.29 of scenarios, 0.08 per attempt with updated safeguards (Tables 5.2.2.3.A and 5.2.2.3.B) | Robustness |
| Feb 5, 2026 | Claude Opus 4.6 system card | 0.29 | Best-of-N browser prompt injection, 389 scenarios, extended thinking, without safeguards, per-attempt attack success rate, lower is better; 2.06 of scenarios (Table 5.2.2.3.A); standard thinking with updated safeguards 0.08 per attempt and 0.77 of scenarios (Table 5.2.2.3.B) | Robustness |
| Nov 24, 2025 | Claude Opus 4.5 system card | 8 | Claude for Chrome internal adaptive evaluation, attack success rate, model only without safeguards, 100 attempts per environment, lower is better (Figure 5.2.2.3.A; 7.6% with old safeguards, 1.4% with improved safeguards) | Robustness |
One lab claim next to the leaderboard
Each score a lab reported for its own model, beside the same model's best score on the leaderboard. Gaps mostly come from different harnesses, effort levels and task subsets. Every benchmark.
| Model | Lab says | Board says | Gap |
|---|---|---|---|
| Claude Opus 4.5 | 8.00% Claude for Chrome internal adaptive evaluation, attack success rate, model only without safeguards, 100 attempts per environment, lower is better (Figure 5.2.2.3.A; 7.6% with old safeguards, 1.4% with improved safeguards) | 1.40% Leaderboard, improved safeguards, Claude for Chrome | −6.6 pts |
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%