Chinese military researchers used OpenAI and Anthropic tools, Reuters says

Chinese military-linked researchers used outputs from OpenAI’s GPT-3.5 and Anthropic’s Claude to train domestic defense systems, even though both companies bar military use. More than 80 Chinese academic papers and patents tied to the People’s Liberation Army used those outputs. The technique at the center of the case is model distillation, a method in which a smaller model learns from the answers or behavior of a larger frontier model rather than from its original training data. Model distillation can make it faster, cheaper and less compute-intensive to build capable systems.

The papers and procurement records pointed to applications in intelligence analysis, decision support, reconnaissance and other battlefield-related tasks. OpenAI and Anthropic both prohibit military use of their services, yet the materials showed researchers were finding workarounds through intermediaries, commercial accounts or other indirect access routes.

OpenAI’s usage policy, published on Oct. 29, 2025, aimed for its tools to be used safely and responsibly while maximizing user control. The company launched OpenAI for Government on June 16, 2025, and on Feb. 28, 2026, later updated on March 2, 2026, it published an agreement with the U.S. Department of War saying the tools would not be used for domestic surveillance of U.S. persons.

Tensions over military AI use were already visible on Jan. 29, 2026, when the Pentagon and Anthropic clashed over the issue.