Army Exhausts Annual AI Token Budget in Six Weeks

The U.S. Army exhausted its annual allocation of AI tokens in approximately six weeks, forcing the service to reimpose usage limits just one month after announcing unlimited access for employees.

Members of the Army's Combat Capabilities Development Command received an email in mid-June 2026 stating that "the Army CIO pool was exhausted of tokens and had to re-establish limits," according to documents obtained by WIRED. The notification came shortly after the Department of Defense announced in May that nearly half of its 3.5 million employees were actively using AI tools at work.

Rapid consumption across the force

The Army uses Ask Sage, a multimodal generative AI platform that provides access to multiple large language models including Alphabet's Gemini, Meta's Llama, and OpenAI's ChatGPT. The platform is accredited for Controlled Unclassified Information and powers the Army's enterprise LLM workspace.

Under the service's annual subscription to an "enterprise pack," the Army had access to 100 million tokens. In the Ask Sage system, a single token represents approximately 3.7 characters of text or image output. Employees received monthly allotments of at least 200,000 tokens and were automatically allocated more if they exceeded their initial quota. Those who signed up but weren't actively using the platform received emails encouraging greater adoption.

"Apparently the whole Army burned through the whole year of tokens for just one service," an Army employee told WIRED, speaking anonymously because they were not authorized to discuss the matter publicly.

Why it matters

The Army's token depletion reveals a fundamental tension in enterprise AI adoption: organizations are aggressively promoting AI usage without fully understanding consumption patterns or establishing sustainable resource management frameworks. As the Defense Department continues expanding AI tools—including developing systems to replace human judgment in civilian casualty assessments—the inability to forecast and manage basic usage metrics raises questions about readiness for more critical AI deployments. The pattern mirrors challenges at major tech companies like Meta and Uber, which have also struggled to control token consumption after encouraging widespread adoption.

Questions about effectiveness

The email to Army personnel noted uncertainty about whether token allocations would be renewed after October 1. It remains unclear whether tokens used by regular DOD employees draw from the same pool as those working with classified or secret information.

One Army employee expressed skepticism about the tools' utility, describing them as unreliable. The employee recounted one instance where a model falsely claimed to have 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," the employee said. "But an unthinking application and use is not going to result in an effective, efficient, and trustworthy rollout."

The Army's experience follows a broader pattern in the technology industry. Meta recently removed its internal leaderboard tracking token usage after encouraging employees to "tokenmaxx," and Instagram head Adam Mosseri proposed capping token usage per engineer. Uber engineers consumed a year's worth of generative AI tokens in just four months, according to Fortune.

Neither the Army, the Department of Defense, nor Ask Sage responded to requests for comment.

These details were first reported by WIRED.