List your company

Biosecurity Misuse Evaluation on BSD Biosecurity

1.3Misuse RateDirect query
-2.19621.4024568.598Jun 6, 2025

Evaluation results

MethodLinks
Direct queryTarget model=GPT-52025.061.3
AdaptiveTarget model=GPT-52025.061.3
Direct queryTarget model=Claude-3.5-Sonnet2025.063
CrescendoTarget model=GPT-52025.0613
PAIRTarget model=GPT-52025.0613.3
Direct queryTarget model=Claude-3.7-Sonnet2025.0615
Adversarial ReasoningTarget model=GPT-52025.0618.7
CrescendoTarget model=Claude-3.5-Sonnet2025.0620.7
AdaptiveTarget model=Claude-3.5-Sonnet2025.0627.3
Direct queryTarget model=o32025.0631.3
AdaptiveTarget model=o32025.0632.7
PAIRTarget model=Claude-3.5-Sonnet2025.0635.3
Decomposition AttackTarget model=Claude-3.5-Sonnet2025.0641.6
Direct queryTarget model=GPT-4o2025.0642
AdaptiveTarget model=GPT-4o2025.0642
Decomposition AttackTarget model=GPT-52025.0645.3
Decomposition Attack (New)Target model=Claude-3.5-Sonnet2025.0646
PAIRTarget model=o32025.0646
Adversarial ReasoningTarget model=Claude-3.5-Sonnet2025.0646.7
Decomposition Attack (New)Target model=GPT-52025.0650.6
Decomposition AttackTarget model=o32025.0652
CrescendoTarget model=Claude-3.7-Sonnet2025.0652.7
Decomposition AttackTarget model=Claude-3.7-Sonnet2025.0652.8
CrescendoTarget model=o32025.0653.3
Adversarial ReasoningTarget model=o32025.0656.7
PAIRTarget model=Claude-3.7-Sonnet2025.0662.7
PAIRTarget model=GPT-4o2025.0664.7
Direct queryTarget model=Gemini-2.5-pro2025.0664.7
Adversarial ReasoningTarget model=Claude-3.7-Sonnet2025.0665.3
CrescendoTarget model=GPT-4o2025.0665.3
Decomposition Attack (New)Target model=Claude-3.7-Sonnet2025.0665.6
AdaptiveTarget model=Claude-3.7-Sonnet2025.0667.3
Decomposition AttackTarget model=GPT-4o2025.0668.4
Decomposition Attack (New)Target model=o32025.0668.8
Adversarial ReasoningTarget model=GPT-4o2025.0670.7
Decomposition Attack (New)Target model=GPT-4o2025.0674
Direct queryTarget model=o3-mini2025.0677.7
Decomposition AttackTarget model=Gemini-2.5-pro2025.0679.3
Decomposition Attack (New)Target model=o3-mini2025.0681.2
Decomposition AttackTarget model=o3-mini2025.0682
Decomposition Attack (New)Target model=Gemini-2.5-pro2025.0682
Adversarial ReasoningTarget model=o3-mini2025.0684
AdaptiveTarget model=o3-mini2025.0684.7
PAIRTarget model=o3-mini2025.0684.7
CrescendoTarget model=Gemini-2.5-pro2025.0686
CrescendoTarget model=o3-mini2025.0686.1
PAIRTarget model=Gemini-2.5-pro2025.0688
AdaptiveTarget model=Gemini-2.5-pro2025.0688.7
Adversarial ReasoningTarget model=Gemini-2.5-pro2025.0688.7