A LinkedIn analysis by Jay Cadmus says Anthropic’s two-year, $200 million ceiling agreement with the U.S. Department of Defense is at risk over unresolved limits on model use in national security settings. The article says the main dispute centers on Anthropic refusing support for autonomous weapons targeting and domestic surveillance of U.S. persons, while Pentagon officials argue deployment choices should be governed by U.S. law rather than vendor policy. It frames the standoff as a precedent-setting test for future frontier-model defense contracts, including accountability, usage rights, and procurement power between government buyers and model providers.

The $200M Question: Who Controls AI Safety in U.S. Defense - LinkedIn