Why replacing Anthropic at the Pentagon could take months

image via New Scientist

March 6, 2026, 7:15 PM

• •The Pentagon deemed Anthropic a 'supply chain risk'.
• •The transition involves replacing Claude with another AI model within six months.
• •Challenges include offboarding, security reviews, and potential automation bias.
• •Anthropic is challenging the designation, while OpenAI steps in to fill the gap.

The Department of Defense is replacing Anthropic's Claude AI on its classified networks due to safety concerns and a disagreement over control. This transition, while quick on a technical level, presents a significant challenge for military personnel who have grown accustomed to using Claude. The replacement of Claude with another AI model, such as OpenAI's, requires careful offboarding and strict security approvals. The move highlights the complexities of integrating and removing AI in sensitive environments and the evolving relationship between tech companies and the government.

Read original article