The Pentagon is pressing top AI companies including OpenAI and Anthropic to deploy their artificial intelligence tools on classified military networks with fewer restrictions than typical civilian use. During a White House event, Pentagon Chief Technology Officer Emil Michael announced plans to make frontier AI capabilities available across all classification levels, according to sources familiar with the matter.

This push represents the latest development in ongoing negotiations between the military and generative AI firms over battlefield applications. While many AI companies have built custom tools for the U.S. military, these are primarily available only on unclassified networks used for administrative purposes. Anthropic currently holds the distinction of being the only AI company with products available in classified settings through third parties, though government use remains subject to the company's usage policies.

Military officials aim to leverage AI's information synthesis capabilities for decision-making in sensitive operations, but researchers warn that AI errors in classified contexts could have deadly consequences. The debate has intensified as OpenAI recently struck a deal allowing military use of ChatGPT on the unclassified genai.mil network, while discussions with Anthropic have proven more contentious due to the company's restrictions on autonomous weapons targeting and domestic surveillance applications.