Anthropic Disrupts Claude Bioweapons Misuse Attempts — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic published its first threat intelligence report of the year, documenting "malicious use" of Claude Haiku, Sonnet, and Opus between December 2025 and August 2026 and highlighting five case studies of biological weapons development attempts — calling biological misuse "one of the most serious risks of frontier AI model."
- The same report named a Russia-linked group (Midnight Blizzard) that used Claude to build self-rewriting malware that evaded security detection, and noted Claude was used by an Iranian propaganda institution and by Chinese firms attempting to distill its capabilities.
- Google disclosed Tuesday that someone attempted to use its Gemini AI to generate a "step-by-step technical guide for synthesizing weaponised biological agents," showing the pattern extends beyond Anthropic's models.
- Senator Bernie Sanders introduced legislation to ban AI superintelligence and pause advanced AI development, saying anyone ignoring scientists' catastrophic-risk warnings "got to be a moron."
- OpenAI chief scientist Jakub Pachocki called for "voluntary slowdowns" until safeguards are established, warning "no one is prepared for the consequences of a continued rapid rise in machine intelligence."
- President Trump rejected AI doomsday concerns, saying he was more worried about the US losing the AI race and being "put in a very bad position."
Why it matters: Anthropic's report pairs concrete misuse cases — Midnight Blizzard's self-rewriting malware, Gemini's weaponized-biology queries, five bioweapons attempts — with Sanders's bill to halt advanced development, crystallizing the split between AI-governance advocates demanding binding international rules and Trump's competitive-race stance that industry cannot self-police.
Ask SkimNews
