Anthropic, the AI company behind the Claude family of models, released a threat‑intelligence report this week that claims it has identified and thwarted several attempts to use its technology for “malicious activity” that could aid the development of biological weapons.
The report lists five case studies in which actors allegedly used Claude to generate step‑by‑step guides for synthesising weaponised pathogens. According to the company, none of these incidents involved its flagship Mythos‑class models, but the disturbing trend prompted the firm to block access to the involved accounts and share the intelligence with law‑enforcement and industry partners.
Anthropic’s findings come amid a broader wave of warnings from AI researchers. In a September 2026 article, safety researcher Jacob Klein said there is “a greater than 10 % chance that AI could kill all humans” within the next decade if safeguards are not tightened. The same week, OpenAI chief scientist Jakub Pachocki called for voluntary slow‑downs while the United States Senate’s front‑loading committee has already tabled legislation to pause advanced AI work.
The report also documents misuse for other sinister purposes, including fake dating apps, hotel Wi‑Fi scams, surveillance aimed at dissidents, and the creation of software for conventional weapons such as drones, firearms and missiles. Anthropic has accused Chinese AI firms of attempting to replicate its Claude capabilities, further heightening concerns about the weaponisation of language models.
Anticropic’s chief threat‑intelligence officer said the situation is “nuanced” and stressed that “the same data can build a vaccine or a cure.” He added that, unlike Hollywood depictions, there is no single person “hunting to build a bioweapon that kills everyone.”
The company also highlighted six separate incidents of Claude being used to develop software for conventional weapons, implying a growing nexus between advanced AI and traditional arms systems.
For more detail, read Anthropic’s September 2026 threat‑intelligence report and watch the accompanying BBC video summarising Senator Sanders’ advocacy for an AI safety treaty.
