Anthropic Report Reveals Global Misuse of Claude AI for Weapons and Espionage

AI company Anthropic has released a 154 page report detailing how various state sponsored and freelance actors used its Claude AI system for weapons development, cyberattacks, and influence operations. The report identifies multiple high risk activities across several continents, including the development of military hardware and the orchestration of disinformation campaigns.

In Russia, Anthropic found that freelance developers used Claude to create software for autonomous kamikaze drone swarms targeting the Donetsk region of Ukraine. Additionally, a Russian state sponsored hacking group known as Midnight Blizzard allegedly used AI to automate cyberattacks against Ukrainian government ministries and military staff, including the takeover of WhatsApp accounts to export private conversations. The AI was reportedly used to autonomously modify malware to evade security detections.

In Yemen, a cell of threat actors in the north used Claude to develop guidance and control software for a guided rocket, a long range ballistic missile, and a hypersonic glide vehicle. While Anthropic stated there is no evidence an operational device was successfully fielded, the group did conduct a test fire of a guided rocket that appeared to fail.

China based actors were also identified in the report. One actor allegedly used Claude to develop electronic warfare software targeting air defenses in Taiwan, while another used the AI to create specifications for an anti torpedo system for the Chinese navy. A third China based actor used the tool to research foreign high power microwave weapons. Separately, a Chinese app studio used Claude to power thousands of AI characters across a network of dating apps to deceive real users.

Influence operations were also a key focus of the report. Anthropic linked a campaign to UAE government officials that targeted the Muslim Brotherhood and UN experts critical of Abu Dhabi's role in the Sudan war. This operation, involving a persona named Deadshot, created fake social media influencers and a front NGO to manipulate public opinion. In France, an advertising agency called LKM Company allegedly used Claude to run a propaganda as a service operation, mass producing political articles across 70 fake news websites in 20 languages.

Other findings include a tech consultant in Mali who used Claude to build a surveillance system called Lakana 360, which was capable of monitoring approximately 25 million SIM cards for the state intelligence service. Anthropic stated that it has banned the accounts associated with these illegal activities and shared information with public and private partners to mitigate risks.