Anthropic blocks possible attempts to use AI for biological weapons

By Laura Cress & Nardine Saad | 11 September 2026

Anthropic, the AI research company behind Claude, said it has identified and disrupted multiple attempts to use its models for “malicious activity” that could facilitate the creation of biological weapons. The company’s first threat‑intelligence report of the year detailed how actors used Claude Haiku, Sonnet and Opus to access information that could aid weaponisation.

Other major AI firms are also stepping up. Google warned that its Gemini was misused to request a “step‑by‑step guide for synthesising weaponised biological agents,” while Anthropic noted its models were employed by Russian cyber‑espionage groups and Iranian propaganda organisations.

Within the past eight months, the company traced incidents ranging from fake dating apps and hotel Wi‑Fi scams to surveillance tools aimed at dissidents. It also recorded six cases where Claude helped create software for conventional weapons such as firearms and drones, and five where parties sought to develop biological weapons.

Jacob Klein, Anthropic’s head of threat intelligence, told the New York Times that while the misuse is “a nuanced situation,” the same capabilities that could build a biological weapon could also be used to create vaccines or cures.

Political responses have already mounted. US Senator Bernie Sanders has called for a pause on advanced AI and a ban on artificial super‑intelligence, while President Donald Trump dismissed the concerns, arguing that falling behind in AI would leave the US vulnerable. The tech industry has published similar reports to demonstrate its safeguards.

Anthropic said it has shared its findings with authorities and industry partners to better detect, prevent and disrupt malicious use of its models. The report underscores that biological misuse is one of the “most serious risks” of frontier AI models, with “catastrophic consequences” if protections fail.