The US Army has exhausted its AI token limit just over a month after deployment.
Despite promises of unlimited supply, the U.S. Army's token pool quickly ran out due to a surge in demand for AI-generated content, forcing the agency to reimpose restrictions.
Just over a month after the U.S. Department of Defense announced that nearly 50% of its 3.5 million employees had begun applying artificial intelligence (AI) to their work, a digital resource crisis has occurred. The U.S. Army Combat Capabilities Development Command (DEVCOM) has issued an urgent notice about running out of AI token limits, urging employees to restrict their use.
AI resource depletion incident at DEVCOM
In May 2026, the U.S. Army Chief of Information Technology announced plans to provide unlimited AI tokens to the force. However, the actual implementation has gone against expectations. By mid-June, the resource pool held by direct management had been depleted, forcing the military to re-establish strict quotas.
Currently, the U.S. Army has decided to maintain its existing token allocation but has not yet determined whether this resource fund will be extended beyond October 1st. An anonymous U.S. Army employee told WIRED magazine that this situation is similar to the military using up an entire year's worth of tokens in just a few weeks for a single AI service.

Ask Sage platform and token consumption mechanism
The platform responsible for this massive outflow is Ask Sage, a multimodal AI generation system certified to handle Controlled Informant (CUI) information. Ask Sage allows users access to leading large language models such as Google Gemini, Meta's Llama, and OpenAI's GPT.
Technically, a token is the basic unit of data that an AI model processes, which can be a word or a part of a character. With Ask Sage, one token is equivalent to approximately 3.7 characters. The U.S. Army currently operates this system under an annual enterprise subscription package with a limit of 100 million tokens.
The main applications of Ask Sage in the Army include:
- Standardize and categorize HR job descriptions.
- Identify and compare the responsibilities and experience required for each position.
- Support procurement and bidding activities at the Office of AI and Digital Technology (CDAO).
Challenges in the Pentagon's Digital Transformation Strategy
The token depletion reveals a flaw in forecasting the US military's AI needs. While administrative operations consumed 100 million tokens in just over a month, actual military campaigns have consumption scales many times larger. For example, during the 38-day Epic Fury campaign in Iran, the US Department of Defense consumed approximately 20 billion tokens per day for combat operations.
Notably, the Pentagon is still pushing ahead with AI deployment despite cost and resource constraints. A recent report indicates that the U.S. Department of Defense is cutting personnel at the Center for Civilian Defense to replace them with AI tools to accelerate the process of assessing casualties in conflict.
Practical effectiveness and skepticism
Not only the military, but also large tech corporations like Meta and Uber are having to tighten their use of AI due to the excessively high cost of tokens. At Meta, the leadership has had to remove the token consumption ranking system and propose a spending limit for each engineer.
Besides the cost issue, the reliability of AI in a military environment is also a major question mark. Some U.S. Army personnel have expressed frustration that AI models sometimes report completing tasks when in fact they haven't. Military experts warn that the mechanical application of AI in federal administration without careful consideration could lead to wasted resources and reduced effectiveness in actual combat operations.


