Controlled trial finds Claude Opus 4 improves bioweapon acquisition plans
Anthropic reported that participants using safeguard-removed Claude 4 models, especially Claude Opus 4, produced materially stronger hypothetical bioweapon acquisition plans than an internet-only control group. It also documented deployed ASL-3 classifier safeguards and a small earlier wet-lab pilot that found no uplift.
0 comments · 0 votes
Sign in to join the discussion →
No comments yet. Start the discussion.
Why it moved the index
The controlled human-uplift result directly increases evidence that a frontier model can lower information barriers for catastrophic biological misuse, while the deployed safeguards and null eight-person wet-lab pilot limit the strength of the update.
Assessment history
-
R1
Toward 44 · confidence 78
Adds a deduplicated September 2025 controlled biorisk-uplift result and its deployed safeguard context.
14 Aug 2026
Share this page
-
DoomBench assesses “Controlled trial finds Claude Opus 4 improves bioweapon acquisition plans” as evidence moving toward doom, with magnitude 44 and confidence 78 out of 100 in the misuse and incidents category.
-
The DoomBench assessment of “Controlled trial finds Claude Opus 4 improves bioweapon acquisition plans” is based on reporting from Anthropic and records the editorial rationale, source quality, attribution, and revision history.
-
DoomBench summarizes “Controlled trial finds Claude Opus 4 improves bioweapon acquisition plans” as follows: Anthropic reported that participants using safeguard-removed Claude 4 models, especially Claude Opus 4, produced materially...
https://www.doombench.com/news/controlled-trial-finds-claude-opus-4-improves-bioweapon-acquisition-plans-2025-09-05