The Pentagon rolled out OpenAI’s ChatGPT Mil and Starshield AI’s Grok for Government from SpaceX across the military Monday, giving personnel two new commercial AI tools for sensitive, unclassified work.
ChatGPT Mil and Grok for Government are each built for “secure, consistent enterprise use” across a department of more than 3 million military and civilian personnel, the Defense Department said. The tools are available through GenAI.mil, the Pentagon’s internal generative AI platform, which launched in December with Google’s Gemini. More than 1.7 million personnel have used the platform since then,according to the department.
The Pentagon says ChatGPT Mil is designed for “document-heavy unclassified work,” including planning, policy, logistics and administration.
Grok for Government includes multiple reasoning modes, customizable workspaces and reusable “playbooks.” The Pentagon said it can be used across “numerous operational contexts,” including acquisition research and supply-chain management.
Trending
The expansion also comes as the Pentagon faces security concerns over how its workforce uses commercial AI. In June, the Defense Counterintelligence and Security Agency warned that unauthorized “shadow AI” tools could create data leaks and other security risks. Congress has separately ordered the Pentagon to assess cybersecurity risks from authorized and unauthorized AI software.
The rollout deepens the Pentagon’s ties with OpenAI and xAI as the department expands its use of commercial AI and wrestles with how much control companies can keep over military uses of their technology.
That fight came to a head with Anthropic, which resisted Pentagon demands that its Claude models be available for all lawful military uses. Anthropic sought safeguards against mass domestic surveillance and fully autonomous weapons.
The Pentagon designated Anthropic a national security supply-chain risk. A federal judge struck down that designation last week, ruling that the government had unlawfully retaliated against the company.
OpenAI took a different path. The company reached a separate agreement with the Pentagon that allows its models to be used for lawful purposes while maintaining limits on domestic surveillance, autonomous weapons and some high-stakes decisions without human involvement.
OpenAI says it can enforce those limits through contractual and technical safeguards, including keeping control of its safety systems and providing its models through the cloud rather than directly on weapons or other devices.
Monday’s announcement does not say whether those same restrictions apply to ChatGPT Mil. It says the tool can be used for planning but does not specify whether that includes operational or military planning or whether it can be used for intelligence analysis, targeting, weapons-related work or decisions involving the use of force.
Asked about those uses, a Pentagon official said tools on GenAI.mil are certified to handle controlled unclassified information at Impact Level 5 but did not address the specific activities.
In May, the Pentagon announced a separate effort to deploy advanced commercial AI models on classified Impact Level 6 and 7 networks. The department said those systems would support operational uses including data synthesis, situational awareness and warfighter decision-making.
Author
