US Army Reinstates AI Token Limits After Troops Exhausted Allowance in Weeks

Full text

The Army previously urged workers to get behind AI, offering them unlimited tokens

Limits have been reinstated by the long-term future is less certain

Iran operation saw token consumption skyrocket

The US Army has reportedly been forced to reinstate limits on GenAI use after workers quickly used the full allowance of tokens early, revealing that even one of the most well-funded agencies in the world can't keep up with AI's unpredictable and rising costs.

In May 2026, it was revealed that the Army would give users unlimited tokens, but by mid-June that token pool had already run dry.

Now an internal email, verified by Wired , confirmed that these limits have had to be reintroduced as a result, with usage renewed at its current level for the time being.

US Army struggling to keep up with AI token spend

Token availability beyond October 2026 remains uncertain, though, with the Army likely worried about rising costs.

"Apparently the whole Army burned through the whole year of tokens for just one service," an unnamed employee told Wired .

As for the intricacies, Wired reports that an annual enterprise pack for Ask Sage, the Army's chosen AI platform, contained 100 million tokens – enough for around 200,000 tokens per employee per month. Users who exhausted their initial allocations were automatically allocated more, though, making the cap virtually pointless.

To put quantities into perspective, the Defense Department reportedly used 20 billion AI tokens per day during its 38-day Operation Epic Fury in Iran.

More broadly, workers have been actively encouraged to use generative AI across the breadth of their roles so that the Army could find out exactly where the tech could seriously improve productivity.

Army aside, this particular case is the perfect example of how aggressive AI pushes might not always consider the impacts of cost., and with many AI vendors now shifting to unpredictable consumption-based or output-based models over flat per-seat subscriptions, it's becoming increasingly harder to predict and allocate budgets.