Technology and Science10 September, 20263 min read
Anthropic Disrupted Five Claude Cases Linked to Biological Weapons Development
This article is an automated aggregation of coverage from news outlets. Our editorials offer in-depth analysis curated by the NewsCord team.
Anthropic blocked five potential bioweapons-related uses of Claude between December 2025 and August 2026. It could not determine researchers' intent, legitimate or malicious, in those cases.
Claude misuse, Anthropic blocks
Anthropic said it disrupted five cases in which researchers used Claude in ways that could have supported biological weapons development, and it said it could not establish whether the researchers intended to cause harm. The company said one example involved a request to Claude for help preparing a grant application for research into the chikungunya virus, and it described the virus as mosquito-borne with debilitating symptoms including severe pain and fever that can last for weeks or months. Anthropic said the cases surfaced between December 2025 and August 2026, and it said it blocked the activity because the potential consequences were too serious to ignore.
Jacob Klein, Anthropic’s head of threat intelligence, told The New York Times that the situation rarely involves obvious declarations of malicious intent and that researchers can present seemingly legitimate scientific goals while pursuing work that raises serious security concerns. Anthropic said it was using the investigation findings to strengthen protections for its services, including blocking accounts linked to the detected activity and introducing new measures to detect and prevent misuse.
Dual-use science and evasions
Anthropic told The New York Times that it could not determine whether the biological research it encountered was legitimate or malicious because valid biological inquiry can also help engineer dangerous pathogens. Jacob Klein said, "You are not seeing someone in a comic book kind of way say, 'Hey, I want to build a biological weapon to kill everybody,'" and he added, "It's an incredibly nuanced situation," as the company described a boundary it said was hard to draw. In one case, Anthropic said a scientist sought Claude’s help writing a grant application for gain-of-function research on chikungunya virus, and it said the work was intended to be performed at a military research institute.
Anthropic said it found attempts to bypass regional restrictions and evade safety systems, and it said researchers continued their work through other channels after Claude blocked sensitive requests. The report also described a reseller platform that routed refused biology prompts toward models with weaker safeguards, and Anthropic said it banned accounts linked to the activity and shared findings with authorities and other AI companies.
What’s at stake next
Anthropic said it tightened controls around high-risk biological queries while also saying it wants researchers to retain access to useful scientific capabilities, framing the safeguards problem as blocking dangerous assistance without crippling legitimate research into diseases, vaccines and treatments. Susan Monarez, a microbiologist and former U.S. public health official, told The New York Times that advanced models could help malicious actors hide dangerous research behind legitimate scientific objectives. The New York Times reported that Anthropic said it erred on the side of caution because the consequences of missing malicious activity could be severe, and it quoted Jacob Klein saying, "It's an incredibly nuanced situation."
Politico reported that Anthropic said foreign scientists, hackers and security services circumvented restrictions to use Claude for dangerous work, including possible bioweapons research, and it said the company uncovered five cases this year where scientists used its models to research dangerous pathogens. El Confidencial said Anthropic concluded that automated security classifiers are no longer sufficient on their own and that access to advanced biological capabilities in AI will require trusted-access programs that verify users’ institutional legitimacy.