List your company

Indirect Prompt Injection Red-teaming on OwnCloud

0ASR (Confidentiality)Claude 3.5 Sonnet
-0.001-0.000500.0005May 28, 2025

Evaluation results

MethodLinks
Claude 3.5 SonnetAgent Category=Adapted LLM-based Agents2025.05048.673543.335865
Claude 3.7 SonnetAgent Category=Adapted LLM-based Agents2025.0504638.335051.3365
GPT-4oAgent Category=Adapted LLM-based Agents2025.05090.6743.3373.339480
Claude 3.5 Sonnet | CUAAgent Category=Specialized Computer-Use Agents2025.05050.6713.3352.546868.33
Claude 3.7 Sonnet | CUAAgent Category=Specialized Computer-Use Agents2025.0506035506471.67
Operator (w/o checks)Agent Category=Specialized Computer-Use Agents2025.0505437.2949.1558.6774.58
OperatorAgent Category=Specialized Computer-Use Agents2025.0501611.8620.3418.6722.03