descriptive claim
In Anthropic's controlled bioweapons-acquisition-planning uplift trials, participants given access to Claude 4 models (especially Claude Opus 4) with safeguards removed produced acquisition plans with substantially higher expert-graded scores and fewer critical failures than a control group with only basic internet access.
desc_bioweapons_uplift_trial_claude4_higher_scores
confidence 0.80
Evidence (1)
supports (1)
- Why do we take LLMs seriously as a potential source of biorisk? direct_measurementweight0.85
locator: Figure 2 and 'Is an LLM's knowledge useful in an applied scenario?' section
“Participants with access to Claude 4 models---especially Claude Opus 4---received much higher scores and developed plans with substantially fewer critical failures compared to the internet-only control group.”