The U.S. Department of Defense says that while the use of AI is rapidly increasing and improving efficiency, fewer than half of government employees have adopted the technology. And the jury is out on whether the product is actually good.
Emil Michael, the Pentagon’s chief technology officer, said at a recent event with conservative think tank Hudson Institute that the use of AI has skyrocketed across the Department, with 80,000 employees using commercial AI tools as of December 2025, and that number growing to 1.5 million as of this month. The Department of Defense has approximately 3.5 million employees, and less than half (43%) use AI.
This is a lot of talk, especially since Elon Musk’s short tenure at the Department of Government Efficiency (DOGE) started with sweeping claims about the efficiency of AI that led to mass layoffs, without much evidence otherwise. A closer look at Michael’s claims suggests that while hiring numbers are surging, they are struggling in ways that reflect broader hiring issues within Fortune 500 companies.
Government use of AI has steadily expanded in recent years, with the Office of Management and Budget last year publishing more than 3,600 ongoing or planned use cases for the technology, a 70% increase over the previous year. In addition to producing congressional reports, these use cases include developing a system for the Federal Bureau of Prisons to assess “the potential for misconduct in newly admitted inmates” and assign security levels to inmates. Court documents reveal that DOGE used ChatGPT to flag federally funded diversity, equity, and inclusion initiatives and rescind a $350,000 grant to replace a museum’s heating, ventilation, and air conditioning (HVAC) system.
The adoption of AI in government agencies snowballed after the creation of DOGE, which effectively replaced the U.S. Digital Service, an Obama administration-era effort to modernize government technology. In 2025, Google and other U.S. technology companies signed an agreement with the U.S. General Services Administration (GSA) to make their tools available to federal agencies at a discount. In May, the Department of Defense announced a series of partnerships with SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, Amazon Web Services, and Oracle for the operational use of their AI tools.
The Department of Defense has been leading the rest of the federal government in integrating AI into its operations for decades. While the government introduced AI efforts to agencies as early as the 1960s, with early experiments in the technology used to solve logistics problems at the Department of Defense, the AI in Government Act of 2020 during the first Trump administration led to the allocation of resources from the GSA and accelerated commercialization of the technology.
The Pentagon has other problems as well. luck Senior contributing columnist Steve Hanke pointed out that the company has never passed an audit. Is leveraging AI really enough to overcome such internal dysfunction?
Defense Department’s victory over bureaucracy
If one of DOGE’s goals was to leverage technology to reduce time spent on bureaucracy, Pentagon officials say they have achieved a victory. Part of the integration of AI tools is the creation of mandatory reports to Congress. According to Michael, using this technology has significantly reduced the time it takes to create these reports. The new defense spending bill increased the number of Department of Defense-mandated reports to Congress from 500 in 2000 to about 1,400 in 2020, according to data from the U.S. Government Accountability Office (GAO).
“I have to report to Congress on this every year,” Michael said at the event. “Put all your documents on it and have it draft a report to Congress. We’ll do it in five hours, something that would otherwise take 200 hours of staffing.”
This isn’t the first time Pentagon officials have praised AI for saving time in producing such reports. Earlier this year, at a Box Federal Summit event, Jacob Glassman, the Pentagon’s assistant secretary of defense for science and technology foundations, said he had asked an understaffed team to use GenAI.mil, the Pentagon’s enterprise AI tool, to create a report for Congress, DefenseScoop reported. Glassman declined to say what reports the tool generated.
“Sure enough, about a week later they came back to me and said, ‘We didn’t just write the report; [but] “This is the best report we’ve written in the last five years,” Glassman said.
The agency did not respond luckThis is a comment request from .
Risks and challenges of government use of AI
Extensive reports on AI adoption in government show slow adoption, full of challenges and stopgaps. The Brookings Institution analyzed federal employment data through the end of 2025, OMB memorandums, and interviews with current and former engineers at eight federal agencies. The conclusion was that five large federal agencies accounted for more than half of the government’s reported AI inventory usage over the past three years. Eleven small federal agencies reported only 2% of their total AI inventory usage in 2025.
Brookings said data tracking AI adoption in the federal workforce is inconsistent, but that “employee capacity constraints, a culture of risk aversion, procurement and funding challenges, and low public trust in AI systems” are slowing implementation efforts. With few resources to train or familiarize staff with AI, agencies often find employees frustrated and ill-equipped to experiment with and use AI.
It’s a similar story in the public sector. The recent Goldman Sachs March 2026 AI Adoption Tracker shows that while AI adoption is increasing, it is not progressing as fast as it would need to become a standard workplace tool. According to the Census Bureau’s Business Trends and Outlook Survey, a whopping 80% of businesses across the economy have not implemented AI, with the actual adoption rate dropping to 19.5% as of May 2026. Adoption is expected to reach 22.7% over the next six months. Businesses with more than 250 employees continue to lead in adoption rates, but at only 36.1%.
Analysts say that while AI is saving workers up to an hour per day, it has not been fully adopted and the productivity gains are small or non-existent. SAP subsidiary WalkMe’s annual Digital Adoption Status Report released in April found that more than half of employees surveyed bypass AI tools and perform tasks manually. Only 9% of employees say they trust technology for complex tasks.
Entering AI also comes with risks. GAO’s March report found that among government and industry experts the agency spoke to, many believed that while AI could make government data easier to cross-reference and more accessible across departments, it could also be used to intentionally or unintentionally misinform the public. Increased access to data also means that agencies, businesses, and other organizations may access it for purposes beyond its intended purpose, such as using tax return information to determine product prices.
The watchdog agency asked OMB to identify privacy-related risks associated with increased use of AI and provide guidance on considerations that agencies should consider when deploying these tools.
The report warns that “without providing this additional information, government agencies are at risk of revealing sensitive information that could have a negative impact on the public.”
