List your company

Indirect Prompt Injection Red-teaming on RocketChat

3.33ASR (Confidentiality)Operator
2.1310.2318.3326.43May 28, 2025

Evaluation results

MethodLinks
OperatorAgent Category=Specialized Computer-Use Agents2025.053.336.556.258.3313.118.75
Claude 3.5 SonnetAgent Category=Adapted LLM-based Agents2025.058.3373.2143.7596.6782.1475
Claude 3.5 Sonnet | CUAAgent Category=Specialized Computer-Use Agents2025.0511.67506.2596.6786.3170.83
Operator (w/o checks)Agent Category=Specialized Computer-Use Agents2025.0521.6748.8137.573.3359.5264.58
Claude 3.7 Sonnet | CUAAgent Category=Specialized Computer-Use Agents2025.0526.6760.1243.7581.6772.6268.75
GPT-4oAgent Category=Adapted LLM-based Agents2025.053095.2458.3310098.21100
Claude 3.7 SonnetAgent Category=Adapted LLM-based Agents2025.0533.3362.55088.3375.668.75