US Army Exhausts Year’s Supply of AI Tokens in Months, Reinstates Usage Limits

A little over a month after the Department of Defense (DOD) announced that nearly half of its 3.5 million employees were using AI at work, members of the Army’s Combat Capabilities Development Command (DEVCOM) received an email informing them that they were burning through tokens and needed to limit use.

“Although the Army CIO announced in May 2026 that they were offering unlimited tokens, by mid-June the Army CIO pool was exhausted of tokens and had to re-establish limits,” the email reads. The email goes on to say that although the Army has chosen to renew token usage at “its current levels,” it’s unclear “if the Army CIO pool will be renewed after 1 Oct.”

The Army uses Ask Sage, a multimodal generative AI platform that provides access to models such as Alphabet’s Gemini, Meta’s Llama, and OpenAI’s ChatGPT. The DOD had an annual subscription for 100,000,000 tokens — each equivalent to about 3.7 characters — and the Army devoured them in a matter of months.

According to an Army employee who spoke to WIRED anonymously, the entire service branch “burned through the whole year of tokens for just one service.” The Defense Department was also using some 20 billion tokens per day during the 38-day Operation Epic Fury in Iran, per Breaking Defense.

The Army had pushed employees to embrace generative AI, allocating at least 200,000 tokens per month per user and automatically topping up those who exhausted their allotment. Employees who were not using their tokens received email reminders to engage more.

But the employee noted that the tools were not particularly reliable, with one model even falsely claiming it had completed a task. “I think there are definitely several aspects of the bureaucracy of the US federal government that these tools might be helpful with. But an unthinking application and use is not going to result in an effective, efficient, and trustworthy rollout.”

The Army’s overconsumption mirrors similar issues at other organizations. After encouraging employees to “tokenmaxx,” Meta removed its internal token usage leaderboards and is now working to clamp down on usage. Uber also saw its engineers burn through a year’s worth of generative AI tokens in just four months.

It remains unclear whether the tokens used for everyday DOD tasks come from the same pool as those for classified work. Meanwhile, the Pentagon continues to lean into AI: the Intercept reported that the DOD has cut staff at the Civilian Protection Center of Excellence, which was responsible for preventing civilian casualties, and is developing an AI tool to replace their assessments.