Anthropic Detects and Blocks AI‑Assisted Biological Weapon Plans
Anthropic announced today that it has identified and disrupted multiple attempts to misuse its Claude model for activities that could contribute to the creation of biological weapons. The firm’s first threat intelligence report of 2026 alleges that actors linked to Russia‑based cyber‑espionage, Iranian propaganda, and other state‑backed groups have used Claude to draft step‑by‑step instructions for synthesising weaponised agents.
The report details more than eight months of malicious use, ranging from fake dating apps and hotel Wifi scams to surveillance of political dissidents. It also highlights four incidents where the model was employed to build software for conventional weapons, including firearms, missiles and armed drones.
None of the cases involved the powerful Mythos‑class models; the only exception was a single instance of distillation—training smaller models from larger ones—for malicious end uses.
“Biological misuse is one of the most serious risks of frontier AI models,” said Jacob Klein, Anthropic’s head of threat intelligence. “The same information that can be used to develop a biological weapon could also be used to develop, for example, a vaccine or a cure for a disease.”
The report has prompted calls for stronger safeguards. US Senator Bernie Sanders has introduced legislation to ban AI super‑intelligence, while the Biden administration has called for a multinational treaty on the safe development of AI. Anthropic says it has shared its findings with authorities and industry partners to improve detection and prevention.
The company’s public threat intelligence indicator is part of a growing trend among AI firms to demonstrate how they monitor and neutralise misuse of their models, following similar disclosures from Google, OpenAI and others.
