GAO Urges Agencies to Rethink AI Spending
Federal officials warn that expiring introductory AI pricing will slow agency adoption and raise costs as token-based billing takes hold.
Updated on Nov. 12, 2026 in Artificial Intelligence
Live Poll
Do you think the federal government is getting better or worse at managing costs?
GAO Chief Scientist Sterling Thomas urged federal agencies to rethink how they deploy artificial intelligence to contain costs. Executive branch agencies have relied heavily on trial periods and discounted pricing from AI companies, and that access is now shifting to paid services.
Why it matters
Token-based pricing bills organizations for the volume of data an AI model processes, so costs can grow substantially when agencies use AI tools at scale. Expiring introductory deals are expected to slow federal AI use over the next eight to 10 months, affecting how agencies budget for and secure these systems across the United States.
Token-based pricing bills organizations for the volume of data an AI model processes, and costs can grow substantially at scale. Agencies are now monitoring whether AI agents retain the information needed to complete tasks, but how much this will cost has not yet been quantified.
The players
Government Accountability Office (GAO)
The GAO is the independent, nonpartisan investigative arm of Congress responsible for auditing federal agency spending and programs.
Sterling Thomas
Sterling Thomas is GAO's chief scientist, who urged federal agencies to rethink AI deployment to contain costs.
Justin Herman
Justin Herman is the federal civilian director at Cognition, an AI company whose work focuses on federal technology adoption.
Nick Marinos
Nick Marinos is GAO's managing director for IT and cybersecurity issues, overseeing the agency's technology audits.
The details
Sterling Thomas, GAO's chief scientist, urged agencies to rethink AI deployment as they move from trial-based access to paid services. Justin Herman, federal civilian director at Cognition, said contracts built solely around consumption tend to increase consumption, while GAO's Nick Marinos warned that automated systems relying on inaccurate data could introduce new vulnerabilities.
Timeline
2002: The Federal Information Security Modernization Act was enacted.
2014: FISMA was amended.
2021: The Office of the National Cyber Director was created, though Congress has yet to update cybersecurity legislation to reflect it.
2023: GAO issued AI governance recommendations.
Next eight to 10 months: Federal AI use is expected to slow as introductory pricing expires.
Roadmap
Congress has not updated the Federal Information Security Modernization Act, enacted in 2002 and amended in 2014, to reflect the 2021 creation of the Office of the National Cyber Director. GAO now warns that AI will intensify existing cybersecurity challenges and speed up cyberattacks, and Congress could revisit FISMA in 2027.
Federal agencies may slow or limit AI tools as paid token-based pricing replaces discounted trials, which can affect the government services citizens rely on. Agencies are also tightening how they monitor and reset AI agents to protect data integrity.
The takeaway
Agencies that plan for usage-based AI costs and data governance now can avoid budget shocks and security gaps later. Treating AI as a metered, security-sensitive resource rather than a free trial is the practical shift this story points to.
What happens next
Federal AI use is expected to slow over the next eight to 10 months as introductory pricing expires, and Congress could revisit FISMA in 2027.
Further reading
For broader coverage of how governments are adopting these tools, see Artificial Intelligence in the Tech section.
Source note: This article includes information reported by Executive Gov.
Live Poll
Do you think the federal government is getting better or worse at managing costs?
