Just weeks after the Department of Defense boasted that nearly half of its 3.5 million employees were using AI at work, the US Army’s promise of unlimited AI tokens has collapsed under the weight of its own success. Members of the Army’s Combat Capabilities Development Command (DEVCOM) received an internal email informing them that the token pool was exhausted and that usage must be limited immediately.
How the unlimited promise unravelled
The Army CIO announced in May 2026 that unlimited AI tokens would be available for all users. But by mid-June, the pool was drained. “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 read. The platform in question is Ask Sage, a multimodal generative AI system that allows users to run different large language models (LLMs), including Alph... The rapid consumption suggests that adoption was far higher than anticipated, or that the token allocation was insufficient from the start.
Why this matters for the military’s AI ambitions
The incident exposes a critical gap between the Pentagon’s public enthusiasm for AI and the practical realities of scaling such tools across a massive workforce. The DOD had earlier highlighted that nearly half of its 3.5 million employees were using AI, framing it as a success story. But the token exhaustion reveals that infrastructure planning may not have kept pace with adoption. For soldiers and civilian staff relying on AI for daily tasks, the sudden limit could slow operations, training, and decision-making.
Timeline of the token crisis
May 2026: Army CIO announces unlimited AI tokens. Mid-June 2026: Token pool exhausted. Limits re-established. Internal email sent to DEVCOM staff warning of the situation. The email also noted that while the Army has chosen to renew token usage at “its current levels,” it is unclear “if the Army CIO pool will be renewed after 1 Oct.” This leaves a cloud of uncertainty over future AI access for the entire Army.
Who is affected and how
DEVCOM staff are the most directly impacted, but the ripple effects could extend across the Army. Engineers, data analysts, logistics planners, and researchers who had integrated Ask Sage into their workflows now face restrictions. The sudden limit may force units to prioritize which tasks get AI support, potentially slowing innovation in combat development and technology testing.
Army CIO’s response and the path forward
The internal email confirmed the renewal of tokens at current levels, but did not specify whether this means a return to unlimited access or a capped system. The ambiguity around the October 1 deadline suggests that budget or capacity constraints may be at play. The Army CIO has not issued a public statement, but the email serves as the de facto official communication. It remains to be seen whether the DOD will allocate additional resources or redesign the token system to handle the demand.
What the token exhaustion really means
The exhaustion is not just a technical glitch—it is a signal that the military’s AI infrastructure is struggling to match its rhetoric. The DOD’s boast of 3.5 million AI users now looks premature when the system cannot sustain that usage. The token model, common in commercial AI platforms, was likely designed to prevent runaway costs, but the Army may have underestimated how quickly its workforce would adopt the tool. This is a classic case of demand outstripping supply in a high-stakes environment.
Confirmed facts vs what remains unclear
Confirmed: The Army CIO announced unlimited tokens in May 2026. The token pool was exhausted by mid-June. DEVCOM staff received an email warning them to limit use. The Army renewed tokens at current levels. It is unclear if the pool will be renewed after October 1. Unclear: The exact number of tokens consumed. Whether the DOD will increase funding. How many employees were affected beyond DEVCOM. Whether the token limit is temporary or permanent.
Ask Sage: The platform at the centre
Ask Sage is a multimodal generative AI platform that allows users to run various LLMs, including Alph... Its adoption across the Army was seen as a step toward modernizing military operations. However, the token exhaustion raises questions about whether the platform’s pricing or capacity model was appropriate for a government-scale deployment. The Army may need to negotiate a new contract or switch to a different provider if demand continues to grow.
Risks and balanced view
While the token exhaustion is a setback, it also reflects genuine enthusiasm for AI within the military. The problem is not that AI is unwanted—it is that it is too popular. Critics may argue that the DOD moved too fast without proper infrastructure planning. Supporters might say this is a growing pain that can be fixed with more investment. The real risk is that if the token issue is not resolved, trust in the Army’s AI initiatives could erode, slowing future adoption.
Wider trend: Government AI scaling challenges
The Army’s token crisis is part of a broader pattern. Governments worldwide are racing to adopt AI, but many struggle with cost, capacity, and governance. The UK’s NHS, for example, faced similar issues with AI tool access. The US federal government has also grappled with cloud computing limits. The lesson is clear: AI adoption at scale requires not just enthusiasm, but robust infrastructure, clear budgeting, and realistic planning.
What Army personnel and contractors should do now
If you are a DEVCOM staff member or Army contractor using Ask Sage, prioritize essential tasks that require AI assistance. Monitor official communications from the Army CIO for updates on token renewal. Consider caching outputs or using offline tools where possible. If your work is critically dependent on AI, raise the issue with your chain of command to ensure your needs are documented for future budget planning.
Future outlook
The Army faces a decision point before October 1. It can either increase token capacity, restructure access tiers, or accept that unlimited AI was never sustainable. The most likely outcome is a hybrid model: limited tokens for routine tasks and priority access for high-impact projects. The DOD may also use this experience to push for a dedicated AI budget line, separate from general IT spending. The token crisis could ultimately lead to a more mature, sustainable AI strategy.
Our Take
The Army’s token exhaustion is a classic tale of ambition meeting reality. The DOD’s pride in having 3.5 million AI users was genuine, but the infrastructure to support them was not ready. This is not a failure of AI itself, but of planning. The good news is that the problem is solvable—with more tokens, better forecasting, and clearer policies. The bad news is that until October 1, thousands of soldiers and civilians are left in limbo. For an institution that prides itself on readiness, this is a lesson in the importance of preparing for success, not just for failure.
Frequently Asked Questions
Why did the US Army run out of AI tokens?
The Army CIO announced unlimited AI tokens in May 2026, but demand from employees using the Ask Sage platform was so high that the token pool was exhausted by mid-June. The system was not designed to handle the rapid adoption.
What is Ask Sage?
Ask Sage is a multimodal generative AI platform used by the US Army. It allows users to run different large language models (LLMs) for tasks like data analysis, report generation, and decision support.
Will the Army restore unlimited AI tokens?
The Army has renewed tokens at “current levels,” but it is unclear if the CIO pool will be renewed after October 1. The future of unlimited access remains uncertain.
How does this affect Army personnel?
DEVCOM staff were told to limit AI use immediately. Other Army employees may also face restrictions, potentially slowing work that depends on AI tools.
