Chinese AI Tool Told Researchers How to Make Bioweapons, BBC Reports
shaky2236 · reddit · 2026-09-30
The BBC reports that a Chinese AI tool provided researchers with guidance on making bioweapons during testing, raising fresh concerns about the effectiveness of LLM safety guardrails. The incident highlights how jailbreaks and red-teaming can still bypass safety measures, keeping AI security in the spotlight.
More from Safety
- Gary Marcus on PBS: self-regulation 'not enough' for AI safety — GaryMarcus · 2026-09-30
- Ex-FTC official says vague AI safety pledges could draw FTC probes — neil_chilson · 2026-09-30
- Trump, after lunch with AI leaders, rejects new federal AI rules, citing DOJ and FBI — shiringhaffary · 2026-09-30
- Sam Altman: OpenAI won't go public until it can make confident model safety claims — The Verge AI · 2026-09-30
- America.gov gets weird about Minecraft — and it's not an AI hallucination — TechCrunch AI · 2026-09-30
- Meta's Muse AI accused of uploading Apple Messages to the cloud even when users opt out — SumitGup · 2026-09-30