Anthropic caught 5 bioweapon attempts; experts say a novel bioweapon within 20 years is near-certain
alexisgallagher · x · 2026-09-12
- Reacting to Ashish Jha's comments on Anthropic's biosafety report (which caught 5 misuse attempts), Lyman Stone argues: the risks are real and severe; it's near-100% odds a terrorist creates a novel bioweapon in the next 20 years even if model progress stopped today; and the only real defense is medical research/production/AI powerful enough to scale treatments and vaccines extremely fast.
- Jha notes we won't get five more warnings before something bad happens.
Related event: NYT Report on Anthropic Halting Bio-Weapons Research Draws Backlash(4 posts)→
More from Companies & People
- Role reversal: universities chase impact metrics while frontier labs fund basic research — IgorCarron · 2026-09-12
- Salesforce quietly ditches Agentforce rebrand after renaming a dozen products — matt_slotnick · 2026-09-12
- Ex-Google researcher: speaking out on AI safety cost him any chance of promotion — Turn_Trout · 2026-09-12
- Alibaba Cloud and Cambricon join PyTorch Foundation board as Platinum members — OkCan8173 · 2026-09-12
- Manifund hires Caroline Ellison, rattling the AI grants community — livgorton · 2026-09-12
- YC founder spotlights four agent startups from his batch, from cloud agents to GitHub rival — ycombinator · 2026-09-12