Anthropic Intercepts AI‑Driven Bio‑Weapon Plans, Highlights Growing Risks

Fri Sep 11 2026 12:11:40 GMT+0000 (Coordinated Universal Time)

Anthropic reveals incidents where its Claude model was used to design biological weapons, prompting calls for tighter AI safeguards.

Anthropic has warned that its Claude models were hijacked by malicious actors to devise biological weapons. In the latest threat‑intelligence report, the company documented five case studies where the model was employed to outline methods for creating dangerous pathogens. The findings echo a broader industry trend, with senators and researchers urging pauses on advanced AI development to curb catastrophic risks. The tech firm also reported dozens of misuse instances ranging from scams to state‑backed surveillance, underlining the need for stronger safeguards and cross‑industry collaboration.

Anthropic Intercepts AI‑Driven Bio‑Weapon Plans, Highlights Growing Risks

Anthropic announced today that it has identified and disrupted several attempts to use its Claude models for the development of biological weapons. The company's most recent threat‑intelligence report, released Thursday, records five concrete instances where malicious actors allegedly leveraged the AI to design or improve pathogenic agents.

This comes amid a mounting chorus of warnings from AI researchers and policymakers about the unintended dangers of frontier language models. In the same month, AI pioneer Jakub Pachocki urged the industry to implement “voluntary slowdowns” until robust safeguards can be deployed, while Senator Bernie Sanders has introduced legislation to pause high‑risk AI research.

The report notes that the misuse incidents spanned a wide range of threats: from fake dating apps and hotel‑wifi scams to surveillance tools aimed at suppressing dissent, and even state‑backed espionage networks. Some cases involved Chinese labs and a Russian‑based hacking group known as Midnight Blizzard, as well as Iranian propaganda institutions, each of whom tried to use Claude to automate code rewriting when detected by security measures.

In a statement, Anthropic’s threat‑intelligence chief Jacob Klein said the situation was “incredibly nuanced.” He added that while the technology can be used to “develop a vaccine or a cure for a disease,” its misuse could “have catastrophic consequences.”

“The same information that can be used to develop a biological weapon could also be used to develop, for example, a vaccine,” Klein explained in an interview with the New York Times. The company said it had blocked all identified scientists who risked enabling such projects, and had shared intelligence with authorities and industry partners where appropriate.

Anthropic’s “threat‑intelligence report” also highlighted that none of the misuse cases involved its most powerful Mythos‑class models, with only one instance of “distillation” (training smaller models from larger ones) noted. The company credited its detection and disruption efforts to ongoing monitoring of Prometheus knowledge graphs, which track model behavior across dozens of commercial and research users.

Following these disclosures, the OpenAI chief scientist and various NGO leaders have called for a multinational treaty to regulate AI's development, emphasizing that “no one is fully prepared” for the consequences of a rapid rise in machine intelligence.

For further context, check the full report at Anthropic Threat Intelligence Report September 2026, and stay informed through the BBC’s own “Tech Decoded” newsletter.