Anthropic Report: "China and Russia Use Claude AI for Missile and Weapons Development"

Anthropic Publishes Threat Intelligence Report

Chinese Research Institutes Attempt "Illegal Distillation"

Russian Hacking Group Utilizes AI in Cyber Attacks

There are growing observations that artificial intelligence (AI) from the United States is being used in regions such as China and Russia to develop weapons, including missiles and drones.

On September 10 (local time), the American AI company Anthropic released a threat intelligence report detailing incidents from the past eight months in which cyber espionage firms and nation-state-backed hacking organizations exploited Anthropic's AI chatbot service, Claude.

For example, in places such as China, Russia, and Yemen, there have been documented cases where Claude was used to develop software (SW) required for the design and development of conventional weapons, including firearms, missiles, armed drones, and bombs. It has also been reported that Claude was used to develop targeting and control systems for these weapons, as well as to collect information and support procurement of supplies related to weapons programs.

Cover of the threat intelligence report released on the 10th (local time) by Anthropic, a U.S. artificial intelligence (AI) company. Anthropic.

One Russia-based hacking group reportedly made extensive use of Claude in most stages of its operations targeting Ukrainian government, military, and diplomatic personnel, including phishing, hotel Wi-Fi hijacking, and WhatsApp account takeovers.

Additionally, during this period, Anthropic said it blocked attempts at so-called ‘illegal distillation’ from seven China-based research institutes. Anthropic specifically named Alibaba, Moonshot AI, DeepSeek, Xiaomi, among others, stating that they tried to extract Claude’s functionalities on a large scale to improve their own AI models. Moonshot AI and DeepSeek were found to have relayed conversations actually entered by their own users into their AI services to Claude, then used Claude’s responses for training their models. In this process, there were also confirmed cases in which users' sensitive information was unknowingly passed on to Claude.

Jacob Klein, Anthropic’s Threat Intelligence Lead, told local media that the advancement of AI is giving rise to new and unforeseen threats. He commented, “Just a year ago, if someone had tried to optimize the capabilities of a drone or missile software, AI would not have been able to perform such tasks effectively from the outset.”

This content was produced with the assistance of AI translation services.