The Department of Defense is seeking to expand the use of artificial intelligence in both unclassified and classified networks, but negotiations with major AI companies have stalled.
Defense officials want access to cutting-edge models without usage restrictions or heavy guardrails. According to Reuters, military officials say AI should be allowed to be deployed in any way as long as it complies with U.S. law.
The move comes after OpenAI announced Monday that it has made a customized version of ChatGPT available through GenAI.mil, the Department of the Army’s AI platform. The platform, launched in December, is used by about 3 million civilians and military personnel and already includes customized versions of tools from xAI and Google’s Gemini.
“We are putting every chip into artificial intelligence as a combat force,” Secretary of the Army Pete Hegseth said in a press release about the platform. “The Department of Defense is leveraging America’s commercial genius to embed generative AI into the daily rhythm of combat.” “AI tools offer endless opportunities to improve efficiency, and we are excited to see the positive impact AI will have across the Department of the Army in the future.”
The OpenAI version of ChatGPT on this platform is designed to assist with daily tasks such as summarizing policy documents, writing reports, and supporting research. But according to Reuters, Pentagon officials are pushing to deploy AI systems at all classification levels, which could open the door to more sensitive uses such as mission planning and weapons targeting.
An anonymous official told Reuters that the Pentagon is “moving toward deploying cutting-edge AI capabilities across all classification levels.”
Anthropic’s models are currently available in some sensitive settings through third-party providers, but there are significant limitations to their use. According to Reuters, Anthropic executives have told military officials that they do not want their systems to be used to target autonomous weapons or conduct domestic surveillance.
Meanwhile, Semafor reported that Anthropic has not agreed to allow its model to be used for “all lawful uses.” At this time, that tool is not available on GenAI.mil.
These negotiations leave AI companies walking a delicate tightrope. However, some employees oppose military use of their systems and worry that it will make it harder to recruit future employees. On the other side is the Department of Defense, which represents large customers and powerful political forces. Semaphore reported that Anthropic’s stance “has drawn the ire of the Pentagon and the White House.”
At the same time, some OpenAI employees have expressed concerns that exiting defense work would give competitors an advantage, Semafor said.
The Department of Defense, OpenAI, Anthropic, Google, and xAI did not immediately respond to Gizmodo’s requests for comment.
