Anthropic changes AI safety policy amid Pentagon pressure
Anthropic has announced a change in its artificial intelligence safety policy, introducing transparency instead of restrictions. This occurred amidst Pentagon demands for greater access to the Claude chatbot.
According to the announcement made by the company on February 23, Anthropic is shifting from strict limitations to transparency mechanisms, such as public risk reports and safety roadmaps. Previously, the company declared that it would halt the training of new models if it couldn’t ensure safety; however, now, according to Chief Scientist Jared Kaplan, this would put them at a disadvantage in challenging market conditions. He emphasized that the demands of the times require the company to remain competitive.
These changes were accompanied by pressure from the U.S. Department of Defense. Earlier, in the summer of 2025, Anthropic signed a contract with the Pentagon for up to $200 million. At that time, some military systems integrated Claude through partners. The use of Claude during the operation to apprehend Venezuelan President Nicolas Maduro in January 2026 only heightened the tension between the company and the Pentagon.
Experts note that changes in safety policy may enhance Anthropic’s competitiveness. However, amid growing pressure from governmental structures, such decisions may have ambiguous consequences for the company and the IT industry in general.
| Date | Event |
|---|---|
| Summer 2025 | Anthropic signed a contract with the Pentagon for up to $200 million. |
| January 2026 | Use of Claude during the operation to apprehend the Venezuelan president. |
| February 23, 2026 | Announced changes in safety policy, meeting at the Pentagon with the Secretary of Defense. |
