What specific biological weapons-related misuse did Anthropic block, and what was the nature of the prohibited research?

Anthropic blocked a request for Claude’s assistance in authoring a grant application involving gain-of-function research on the chikungunya virus—specifically aimed at enhancing its transmissibility and immune evasion properties—which could make the virus more harmful and thus pose dual-use risks for biological weapons development, even though such research could also support vaccine or treatment development.

How does Anthropic justify its decision to disclose these misuse cases, and what broader concern does it highlight about AI model capabilities?

Anthropic states it published the report because it believes it has a responsibility to disclose malicious misuse of its services, emphasizing that as AI models become increasingly capable, their risks will increase unless AI developers and society’s defenders proactively make them safer—a warning underscored by findings that newer models like Claude Fable or Mythos-class could meaningfully assist in complex scientific research, including dangerous biological research, unlike older models such as Claude Opus 4 and Sonnet 4.5, which were 'well below the threshold' for such assistance.

Anthropic Says It Blocked AI Misuse That Could Support Biological Weapons

Sep 11, 2026

Anthropic reported blocking attempts to use its AI models for biological weapons research, citing gain-of-function studies on the chikungunya virus and implementing stricter safeguards on newer models.

apnews

2026-09-11

Anthropic Says It Blocked AI Misuse That Could Support Biological Weapons

2026-09-18

Anthropic Quietly Sets Up Biology Lab as It Ramps AI Drug Program

Anthropic Says It Blocked AI Misuse That Could Support Biological Weapons

FAQ