Anthropic Blocks Attempts to Use AI for Potential Bioweapon Development

A growing concern over AI misuse has emerged after Anthropic found cases involving research linked to potential biological weapons development. The company has blocked several accounts after identifying AI use that could support such research, according to a report published on September 10. The Claude maker said it had disrupted dozens of potentially malicious uses of its models since December 2025. The latest report describes five cases involving activities that could support biological weapons development. Anthropic said these cases highlight a growing challenge as advanced AI models become more capable in sensitive scientific fields.

The new report also details attempts by users to bypass Anthropic’s safety controls. Some users reportedly “circumvented” access restrictions designed to block users from certain regions. In addition, some users tried to hide the purpose of their research to avoid the company’s safeguards. However, Anthropic has stressed that some flagged research had legitimate scientific value, which can make harmful intent difficult to establish. The cases therefore show the growing challenge of detecting AI-enabled biological risks while allowing legitimate scientific research to continue.

AI Misuse Concerns Spread From China’s Surveillance to Yemen’s Weapons Work

AI has become an important tool for drug development, scientific research and medical discovery. It can help researchers understand diseases, analyze large amounts of data and develop new treatments. However, the same capabilities can create risks when people use them for harmful purposes. Anthropic has warned that bad actors could present sensitive research as legitimate scientific work, making their intentions harder to identify. The company said recent models, including Claude Fable 5, have stronger safeguards that restrict access to a wide range of dual-use biological research queries.

Anthropic has also reported suspected efforts involving China and northern Yemen. According to the company, China-linked operators had attempted to use Claude for surveillance and intelligence-related activities involving Uyghurs and other religious communities. Anthropic also raised concerns about Moonshot and DeepSeek, alleging that some requests had been routed to Claude or that its responses had been used for AI model distillation. In northern Yemen, Anthropic said some actors had tried to use Claude for work involving missile guidance software, although it did not identify the group and media reports have linked the case to Iran-backed Houthi forces. Anthropic did not say Claude had built or launched a missile. Instead, the cases show how advanced AI can extend into surveillance and weapons-related work, highlighting the growing need for stronger safeguards as AI models become more capable.

Bird Flu And Chikungunya Research Raise Biological Safety Concerns

One case involved a researcher outside the United States who accessed Claude from an unsupported region while studying highly pathogenic avian influenza, commonly known as bird flu. According to Anthropic, the research examined how the virus could adapt to mammals. The company described the work as clearly dual-use because the same scientific knowledge can help researchers identify emerging virus variants with pandemic potential. However, that information could also provide insights that might make a dangerous pathogen more harmful.

Another case involved a request to draft a grant application for gain-of-function research on chikungunya. Anthropic said the proposed research focused on the virus’s transmissibility and ability to evade immune responses. In addition, the report flagged research involving orthopoxviruses, a group that includes smallpox and mpox, as well as queries involving venom toxins. As AI capabilities expand, stronger safeguards will be essential to ensure scientific progress does not create new security risks.

Web Resources On Anthropic Blocks AI Bioweapon Attempts

1. CNN : Anthropic says it blocked possible attempts to use AI to develop bioweapons.
2. ABC News: Anthropic says it blocked potential AI bioweapon misuse.
3. Fox Business : Anthropic says it blocked possible efforts to use AI for biological weapons development, Iran-linked cases.