{"slug": "us-army-reimposes-ai-limits-weeks-after-burning-through-unlimited-token-pool", "title": "US Army Reimposes AI Limits Weeks After Burning Through 'Unlimited' Token Pool", "summary": "The US Army has reinstated limits on generative AI use after an 'unlimited' token pool was exhausted within weeks, according to an internal Army email. The Army Chief Information Officer's shared 100 million-token pool was depleted by mid-June, forcing officials to restore restrictions shortly after announcing unrestricted access in May 2026. The rapid depletion highlights the challenge of scaling AI adoption across the Department of Defense, where nearly half of 3.5 million personnel use generative AI tools.", "body_md": "# US Army Reimposes AI Limits Weeks After Burning Through 'Unlimited' Token Pool\n\n## Shared AI token pool ran dry within weeks as Army employees rapidly increased usage.\n\nThe US Army has reinstated limits on employees' use of generative artificial intelligence just weeks after an 'unlimited' access programme burned through its shared token pool, according to an internal Army email, highlighting how quickly large-scale AI adoption can outstrip available computing resources.\n\nThe email said the Army Chief Information Officer's token pool was depleted by mid-June, forcing officials to restore limits only weeks after announcing unrestricted access. The move comes as the [Pentagon expands the use of AI](https://www.ibtimes.co.uk/spacex-expands-ai-computing-pentagon-deal-1809890) across the Department of Defense after saying nearly half of its 3.5 million personnel already use generative AI tools.\n\n## Army Restored Limits Weeks After Offering Unlimited Access\n\nAccording to the internal email, the Army CIO's promise of unlimited AI tokens was short-lived.\n\n'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 said.\n\nThe message added that token availability has been renewed at current levels, but officials remain uncertain whether the shared pool will be replenished after 1 October.\n\nNeither the US Army nor the Department of Defense responded to requests for comment.\n\n## Ask Sage Rollout Drove Heavy AI Adoption\n\nThe Army's rollout centres on Ask Sage, a secure enterprise platform that provides access to several large language models, including OpenAI's ChatGPT, Google's Gemini and [Meta's Llama](https://www.ibtimes.co.uk/enterprise-ai-budgeted-management-1806370).\n\nThe platform is accredited to process Controlled Unclassified Information and supports acquisitions, personnel management and other administrative functions across the Army.\n\nEmployees initially received monthly allocations of at least 200,000 tokens. Those who exhausted their allowance automatically received additional capacity, while staff who rarely used the platform received reminders encouraging broader adoption.\n\nThe Army's enterprise subscription reportedly provided access to a shared 100 million-token pool. Internal documents show each token represents roughly 3.7 characters of generated output.\n\nOne employee familiar with the programme said demand accelerated rapidly once restrictions were removed, exhausting the shared token pool within weeks despite expectations it would last much longer.\n\n'Apparently the whole Army burned through the whole year of tokens for just one service,' the employee said.\n\n## Pentagon Pushes Ahead With Wider AI Expansion\n\nThe Army's experience comes as the Pentagon continues expanding generative AI across military and civilian operations.\n\nAI is being used to support planning, acquisitions, logistics and administrative work as part of a broader effort to modernise defence operations.\n\nSeparate defence reporting has estimated that the Department of Defense consumed roughly 20 billion AI tokens per day during the 38-day Operation Epic Fury in Iran, illustrating the level of computing resources required as AI becomes integrated into military activities.\n\nThe Pentagon is also developing AI-powered tools designed to support operational assessments and other decision-support functions.\n\n## Unlimited Access Met Real-World Limits\n\nThe Army's experience shows one of the early operational challenges facing governments deploying [generative AI at scale](https://www.ibtimes.co.uk/chatgpt-electricity-demand-ai-1806338).\n\nOfficials encouraged broad adoption by automatically increasing token allocations when employees exceeded their monthly allowance and prompting inactive users to make greater use of the platform. Within weeks, however, unrestricted access had been replaced by renewed limits after the shared token pool was exhausted.\n\nRather than indicating a retreat from artificial intelligence, the Army's decision underlines a growing reality for large organisations: demand for generative AI is rising faster than the infrastructure and computing resources needed to sustain unrestricted use.\n\n© Copyright IBTimes 2025. All rights reserved.", "url": "https://wpnews.pro/news/us-army-reimposes-ai-limits-weeks-after-burning-through-unlimited-token-pool", "canonical_source": "https://www.ibtimes.co.uk/us-army-ai-token-shortage-unlimited-access-1810129", "published_at": "2026-07-22 21:12:53+00:00", "updated_at": "2026-07-22 21:39:38.110945+00:00", "lang": "en", "topics": ["artificial-intelligence", "generative-ai", "ai-policy", "ai-infrastructure", "large-language-models"], "entities": ["US Army", "Army Chief Information Officer", "Ask Sage", "OpenAI", "Google", "Meta", "Department of Defense", "Pentagon"], "alternates": {"html": "https://wpnews.pro/news/us-army-reimposes-ai-limits-weeks-after-burning-through-unlimited-token-pool", "markdown": "https://wpnews.pro/news/us-army-reimposes-ai-limits-weeks-after-burning-through-unlimited-token-pool.md", "text": "https://wpnews.pro/news/us-army-reimposes-ai-limits-weeks-after-burning-through-unlimited-token-pool.txt", "jsonld": "https://wpnews.pro/news/us-army-reimposes-ai-limits-weeks-after-burning-through-unlimited-token-pool.jsonld"}}