Overview
- Anthropic disclosed on Thursday that its threat intelligence team identified and disrupted dozens of misuse cases from December 2025 to August 2026, including five that involved biological research the company judged could support weapons development.
- One cited case involved a request to help draft a grant for gain‑of‑function work on the chikungunya virus and several others concerned bird flu, orthopoxviruses, and toxin research, all of which Anthropic says it blocked because intent could not be reliably determined.
- The company said it banned implicated accounts, worked with partners to take down reseller and relay networks used to evade regional restrictions, and shared intelligence with government agencies and other AI labs.
- Anthropic reported none of the documented cases used its most recent Claude Fable or Mythos‑class models except for one instance of illicit distillation, and it said it has applied stronger restrictions to those newer models.
- The publication of the 154‑page misuse report comes days after researcher Jacob Coxon resigned with stark warnings about runaway AI and follows Anthropic's recent IPO preparations that expanded its underwriting group, prompting renewed calls for oversight and mandatory incident reporting.