Anthropic Reports Blocking Attempts to Use Claude for Biological Weapons and Military Drones
Anthropic reported blocking multiple attempts to misuse its Claude AI model for developing biological weapons and autonomous military drones, in operations linked to Russia, China, Iran, and Yemen. The company also documented nine disinformation campaigns using Claude to manage fake profiles and spread false content across six continents.
What happened
Anthropic released a 154-page report in 2026 describing the most serious incidents in which actors attempted to misuse its Claude models or succeeded in misusing them. The company stated it had detected and thwarted several attempts to use its AI models to develop biological weapons.
The report detailed six cases of military misuse, including three in China, two in Russia, and one in Yemen. Anthropic reported blocking a Russian operation attempting to develop an autonomous swarm of military drones. Actors linked to Russian cyber espionage campaigns and an Iranian propaganda institution also used Claude models, according to the document.
The report described five case studies in which actors used models in ways that could support biological weapons development. It also documented nine influence operations from Russia, Iran, Turkey, the Persian Gulf region, South Asia, Africa, and Europe targeting audiences on six continents, using Claude to plan entire campaigns, generate content, and manage fake profiles and disinformation websites.
Reported impact
Anthropic stated that misuse in the biological area is one of the most serious risks posed by advanced AI models. Without proper security measures, such capabilities could have catastrophic consequences, according to the report's authors.
Content generated through these operations appeared in media outlets associated with the Russian government, including Sputnik Moldova, RIA Novosti, Sputnik en Español, Sputnik Africa, and RT.
Index analysis
The incidents represent attempts to use a deployed commercial AI system for weapons development and military operations. The actors sought to circumvent or exploit Claude's safeguards to access information and capabilities for biological weapons, autonomous weapons systems, and military support applications.
Anthropic's ability to detect and prevent these attempts suggests either that the actors' efforts were unsuccessful in overcoming the model's safety measures, or that the company has monitoring mechanisms in place to identify and interrupt such misuse patterns.
Governance controls
Anthropic reported that its teams investigated and blocked the activity of multiple actors attempting to use Claude systems for weapons development and military intelligence gathering. The company's actions prevented the completion of these objectives.
Anthropic disclosed these incidents in a formal public report, indicating a governance approach that includes transparency about detected misuse attempts and security responses.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.