Less than a month after the Department of Defense claimed nearly half its workforce was using artificial intelligence, the Army’s Combat Capabilities Development Command received an order to stop. Members of the unit were told they were consuming tokens too quickly and had to limit their usage.
An internal email stated that although the Army CIO announced unlimited tokens in May 2026, the pool ran out by mid-June. Limits were re-imposed immediately. The message noted that while token usage would continue at current levels, it was unclear if the Army CIO pool would be renewed after 1 Oct.
The service uses Ask Sage, a multimodal platform allowing access to various large language models, including Alphabet’s Gemini, Meta’s Llama, and OpenAI’s ChatGPT. One Army employee told WIRED that the entire service burned through a year’s supply of tokens in a single period. The soldier spoke anonymously because they were not authorised to talk to the press.
How the tool works
Ask Sage powers the Army’s enterprise LLM workspace and is accredited for Controlled Unclassified Information. The DOD’s Chief Digital and AI Office uses the system for acquisitions. The Army website lists tasks such as reclassifying personnel descriptions, which involves defining and aligning job duties, experience and backgrounds.
Employees received an allotment of at least 200,000 tokens per month. If they used them up, they were automatically allocated more. People who signed up for Ask Sage but did not use it regularly received emails encouraging them to consume their allocated tokens.
The Army had access to 100,000,000 tokens as part of an annual subscription to an enterprise pack. Tokens measure output, either in text or image. For Ask Sage, a single token equals about 3.7 characters. During the 38-day Operation Epic Fury in Iran, the Defense Department burned through some 20 billion tokens per day, according to Breaking Defense.
The Army and DOD did not reply to requests for comment. Neither did Ask Sage.
Unclear boundaries
It is unclear if tokens used by regular DOD employees come from the same pool as those using AI tools on classified or secret information. This has not stopped the Defense Department from emphasising AI. On Monday, the Intercept reported that the Pentagon continued to lean into AI tools and cut staff at the Civilian Protection Center of Excellence. Those jobs involved preventing civilian casualties in conflict zones. Instead, the DOD is developing an AI tool to speed up assessments that the Center’s staff would normally make.
Other companies face the same issue
The Army is not the first eager adopter of generative AI to rethink near unlimited use. After encouraging employees to “tokenmaxx”, Meta quietly took down its leaderboard tracking token usage and is now trying to curb use. Last week, Adam Mosseri, head of Instagram at Meta, floated the idea of capping token use per engineer at the company. According to reporting from Fortune, Uber also saw its engineers burn through a year’s worth of generative AI tokens in merely four months.
What it means
The Army employee says they have not found the generative AI tools particularly useful for their work. When used, the tools proved unreliable. One model asserted it had completed a task it had not. The soldier believes there are aspects of US federal bureaucracy where these tools might be helpful. They argued that unthinking application and use will not result in an effective, efficient, and trustworthy rollout.
