Anthropic Report Reveals Widespread Misuse of Claude AI for Weapons and Cyberattacks

AI laboratory Anthropic released a report on Thursday detailing how users have exploited its chatbot, Claude, for activities including the development of weapons, electronic espionage, and fraud. The company stated that it disrupted these activities in progress, though it admitted that safeguards blocked some, but not all, requests.

According to the report, users in China used Claude to develop electronic warfare software to simulate radar interference against Taiwanese targets and to create fire control software for an anti torpedo system for the Chinese Navy. Another user in China allegedly used the tool to research foreign high power microwave weapons and identify supply chains for reverse engineering. The Chinese Ministry of Foreign Affairs denied knowledge of the report and opposed what it called the distortion of facts.

Other military applications included a group in northern Yemen using the AI to develop software for guided rockets and ballistic missiles with ranges exceeding 2,000 kilometers. In Russia, independent agents reportedly used Claude to develop software for autonomous attack drone swarms, while a Russian procurement manager used the tool to find intermediaries in China and Hong Kong to bypass sanctions on European military components.

Beyond conventional weapons, the report highlighted cyberattacks and biological threats. A center left source notes that Russian state sponsored hackers known as Midnight Blizzard used Claude for reconnaissance against Ukrainian and European government networks, while the cybercriminal group ShinyHunters used the tool throughout its hacking and extortion campaigns. Anthropic also detailed five attempts to use its models to support the development of biological weapons, including requests to modify the chikungunya virus.