According to DefenseScoop and The Next Web, the Department of Defense has begun using generative AI to draft reports mandated by Congress, making tools available through a platform called GenAI.mil since December 2025, initially using Google's Gemini on unclassified networks. DefenseScoop reports Pentagon CTO Emil Michael told a Hudson Institute forum that automating a single congressional report could reduce an estimated 200 hours of staff time to about five hours, quoting Michael as saying: "I have to report to Congress every year on this thing. Let me load all the papers onto it and have it draft me a congressional report that would otherwise take 200 hours of staffing time and do it in five hours." DefenseScoop and The Next Web report the department now counts 1.5 million daily users of GenAI.mil out of 3.5 million DoD personnel. Editorial analysis: This public account illustrates adoption at scale inside a major government department and raises operational and governance questions for practitioners and overseers.
What happened
According to DefenseScoop (Jon Harper, June 12, 2026) and The Next Web (June 15, 2026), the Department of Defense has started using generative AI to draft reports that are required by Congress. DefenseScoop reports the department made generative-AI tools available to service branches through a platform called GenAI.mil since December 2025, beginning with Google's Gemini on unclassified networks. DefenseScoop reports Pentagon CTO Emil Michael told a Hudson Institute forum that automating a single congressional report could cut roughly 200 hours of staff work to about five hours, quoting Michael: "More and more people are like, 'Oh my God, I could write a job description.' I mean, very simple things to more exquisite things. 'I have to report to Congress every year on this thing. Let me load all the papers onto it and have it draft me a congressional report that would otherwise take 200 hours of staffing time and do it in five hours.'" DefenseScoop and The Next Web both report the department now counts 1.5 million daily users of GenAI.mil out of 3.5 million DoD personnel, up from 80,000 at launch in December 2025.
Technical details
Editorial analysis - technical context: Public reporting names Google's Gemini on unclassified networks as the initial model family, with OpenAI's ChatGPT and xAI's Grok subsequently added. Companies and agencies adopting hosted LLMs for sensitive-document summarization typically confront three technical areas: secure data ingestion and classification, prompt and template engineering to reduce hallucination, and auditable logs for provenance and versioning. Those patterns are relevant to any large-scale rollout that moves raw or aggregated internal documents through a generative pipeline.
Context and significance
The reporting is notable because it documents generative-AI use at scale inside a national security bureaucracy. For practitioners, deployments of this size create operational trade-offs between productivity gains and risk controls: access management, data leakage, model calibration, and red-team testing become central engineering problems. The public quote about compressing 200 hours into five hours underscores why organizations adopt generative tools, while the reported 1.5 million user figure signals broad internal availability rather than a small pilot. The Next Web notes the Pentagon has not disclosed error rates, accuracy metrics, or any internal assessment of the quality of GenAI.mil outputs.
What to watch
Observers will look for follow-up disclosures on governance and controls: audit trails, data retention and deletion policies, whether sensitive data enters model contexts, and results from any internal red-teaming or external oversight. For practitioners evaluating similar integrations, monitoring how provenance and human-in-the-loop checks are implemented will be an early indicator of operational maturity.
Scoring Rationale #
This is a notable, practitioner-relevant report documenting large-scale generative-AI adoption inside a major government agency. The combination of scale (1.5 million users) and use for legally mandated congressional reports makes governance and reliability issues immediately relevant to engineers and compliance teams.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.