The U.S. Department of Defense is urging leading artificial intelligence companies, including OpenAI and Anthropic, to make their AI tools available on classified military networks with fewer restrictions than those typically imposed on commercial users. The push reflects the Pentagon’s growing urgency to integrate advanced AI systems into sensitive defense operations.
As per the Reuters report, during a White House event on Tuesday, Pentagon Chief Technology Officer Emil Michael told technology executives that the military aims to deploy AI models across both unclassified and classified domains, according to two people familiar with the remarks. A defense official said the Pentagon is “moving to deploy frontier AI capabilities across all classification levels.”
The effort marks the latest phase in negotiations between the U.S. military and major AI developers over how generative AI will be used in future conflicts. Those conflicts are increasingly shaped by autonomous drone swarms, robotics, and cyber operations, where rapid data synthesis and decision support are viewed as strategic advantages.
Tension Over Restrictions and Safeguards
The Pentagon’s request is likely to intensify an already contentious debate between military officials and AI companies. While defense leaders want flexibility to deploy commercial AI systems as long as they comply with U.S. law, AI developers have built safeguards into their models and imposed usage policies to limit potential harm.
Most AI tools currently used by the military operate only on unclassified networks, which handle administrative and non-sensitive tasks. Classified networks, by contrast, support activities such as mission planning, intelligence analysis, and weapons targeting. Deploying AI chatbots in such environments raises higher stakes, as errors or fabricated outputs could have severe consequences.
AI researchers have warned that while large language models can synthesize vast amounts of information, they are prone to mistakes and hallucinations. In classified or combat-related settings, such failures could lead to faulty decisions with potentially deadly outcomes.
Pentagon officials have expressed frustration with restrictions they view as overly limiting. They argue that the military should be able to adapt commercial AI tools to operational needs without being constrained by policies designed for civilian use.
Current Deployments and Industry Deals
So far, only one major AI company, Anthropic, has made its technology available in classified settings, and even then through third-party arrangements that still enforce the company’s usage policies. Other firms have resisted broader classified deployment.
This week, OpenAI reached an agreement with the Pentagon allowing its tools, including ChatGPT, to be used on an unclassified Defense Department network known as genai.mil. The platform has been rolled out to more than 3 million Defense Department employees. As part of the deal, OpenAI agreed to remove several standard user restrictions, though some guardrails remain in place.
Google, part of Alphabet, and Elon Musk’s xAI have previously struck similar agreements with the U.S. government for unclassified use. In a statement, OpenAI said the current agreement applies only to unclassified deployment and that any expansion into classified environments would require a new or revised contract.
A Broader Strategic Shift
The Pentagon’s push underscores how central AI has become to U.S. defense planning. As military leaders seek to accelerate adoption, they are increasingly challenging the boundaries set by private companies. Whether AI developers will relax restrictions for classified use remains uncertain, but the negotiations signal a deeper alignment, and potential friction, between Silicon Valley and the U.S. national security apparatus.
