Anthropic says it blocked AI-assisted bioweapons research attempts

Anthropic published a 154-page threat intelligence report on Thursday revealing it disrupted five cases in which scientists used its Claude AI models for biological research that could support weapons development, alongside cyberattacks, surveillance operations, and conventional weapons work tied to Russia, China, Iran, and Yemen.yahoo+1

The report, covering activity detected between December 2025 and August 2026, represents what the Washington Examiner described as the AI industry's first known admission that biological weapons development overlaps with AI misuse. Anthropic said it could not determine whether the researchers intended harm but chose to act because the potential consequences were too severe to ignore.nytimes+1

Dual-Use Science and a Military Lab

The most detailed case involved a scientist who in May sought Claude's help drafting a grant application for gain-of-function research on the chikungunya virus, a mosquito-borne pathogen that causes prolonged joint pain. The proposed work aimed to engineer mutations making the virus more harmful through repeated infection of live animals, and was intended to be carried out at a military research institute. A second case involved weeks of Claude-assisted planning for experiments on highly pathogenic avian influenza focused on mammalian adaptation. Three additional cases involved orthopoxviruses, venom-related compounds, and toxins.interestingengineering+2

Jacob Klein, Anthropic's head of threat intelligence, told The New York Times that detection is inherently difficult. "You are not seeing someone in a comic book kind of way say, 'Hey, I want to build a biological weapon to kill everybody,'" he said. "It's an incredibly nuanced situation."nytimes

Anthropic noted that older Claude models showed limited ability to assist with sophisticated biological research, but current systems can complete complex scientific tasks — a shift the company said changes its risk calculus.yahoo

State-Linked Actors Across Multiple Domains

Beyond biological threats, the report documents a suspected Russian espionage group using Claude to automate cyberattacks against Ukrainian government targets and European diplomatic organizations. Chinese and Iranian government-aligned actors used Claude to surveil diaspora communities and dissidents. Anthropic also identified six cases of conventional weapons development, including a Yemen-based cell that test-fired a guided rocket and returned to Claude within hours to analyze why it failed.qz+1

Industry Alarm

Andrew Weber, a former U.S. Assistant Secretary of Defense who reviewed the report before publication, called the findings "chilling examples of state-sponsored biological weapons developers tapping into the rapidly advancing capabilities" of leading AI models. He urged that Russia, China, and North Korea be denied access to the most capable systems.yahoo+1

Anthropic banned all accounts linked to the identified activity, dismantled relay networks used to circumvent regional access controls, and shared its findings with governments and other AI companies. The company also disclosed that bio-risk filters were inadvertently disabled on contractor traffic from May 2025 to April 2026, exposing roughly 133 million exchanges, though a retrospective review found no confirmed bioweapons uplift from the gap.interestingengineering+2