The Pentagon is pressing leading artificial intelligence (AI) companies to make their tools available on classified military networks with fewer user restrictions, as Washington explores the broader use of AI in US defense.
Reuters reported on Thursday, Feb. 12, citing two people familiar with the matter, that Pentagon Chief Technology Officer Emil Michael told technology executives at a White House event on Tuesday that the US military aims to deploy advanced AI models across both unclassified and classified systems.
The Pentagon is “moving to deploy frontier AI capabilities across all classification levels,” an unnamed official told Reuters.
Negotiations over AI use in warfare
The talks mark the latest stage in negotiations between the US Defense Department and top generative AI firms over how such technologies could be used in future conflicts, where autonomous systems, drone swarms and cyber operations already play a growing role.
Most AI tools currently developed for the US military operate only on unclassified networks, typically used for administrative functions. Anthropic is the only major AI company whose products are available in classified settings through third parties, though the government remains bound by the firm’s usage policies, as per Reuters.
Classified networks are used for sensitive activities, including mission planning and weapons targeting. Reuters said it could not determine when or how the Pentagon plans to deploy AI chatbots on such systems.
Military officials hope to use AI to synthesize large volumes of information to support decision-making. However, researchers have warned that generative AI models can produce inaccurate or fabricated outputs, which could pose significant risks in classified environments.
Agreements with OpenAI, Anthropic
This week, OpenAI agreed to allow the Pentagon to use its tools, including ChatGPT, on an unclassified Defense Department network known as genai.mil, which has been rolled out to more than three million employees. As part of the deal, OpenAI eased several user restrictions, though some safeguards remain in place.
Alphabet’s Google and Elon Musk’s xAI have previously signed similar agreements, as per Reuters.
OpenAI said the new arrangement applies only to unclassified systems and that any expansion to classified networks would require a separate agreement.
Anthropic, whose products include the chatbot Claude, has previously told military officials it does not want its technology used for autonomous weapons targeting or domestic surveillance. The company said it remains in discussions with the Defense Department regarding future cooperation.
US President Donald Trump ordered the Defense Department to rename itself the Department of War in late summer 2025, a change that would require congressional approval.