In This Article
Key Takeaways
- Through its OneGov strategy, GSA negotiated government-wide deals offering ChatGPT Enterprise and Claude for $1 per agency for a year, and Gemini for Government for 47 cents per agency.
- OpenAI, Anthropic, and Google were all added to the GSA Multiple Award Schedule, the pre-negotiated vehicle agencies buy from.
- The nominal prices are a trial: after the first year, agencies must move to a paid agreement or let access lapse, and post-trial pricing varies by vendor.
- Cheap access to a model is not the hard part of federal AI — accreditation, data boundaries, and audit are, and those are unchanged by the price tag.
What GSA actually negotiated
The number that got attention was the price: about a dollar. Under its OneGov strategy, the General Services Administration cut government-wide deals that let participating federal agencies use OpenAI's ChatGPT Enterprise for $1 per agency for a year and Anthropic's Claude for $1 across all three branches, with Google's Gemini for Government priced at 47 cents per agency for a year, an offer Google said runs through 2026.
OneGov is the mechanism, not a giveaway for its own sake. It is a GSA approach to negotiating volume deals with technology vendors on behalf of the whole government, and per GSA's own announcement, the Anthropic agreement covers the federal civilian executive branch plus the legislative and judicial branches. All three vendors were also added to the GSA Multiple Award Schedule — the pre-negotiated contract vehicle agencies routinely buy from.
Why a dollar
The pricing is a customer-acquisition play aimed at a single, sticky market. Getting a model into an agency's daily workflow for a year — with the security and enterprise features that federal use requires — is how a vendor becomes the default the following year, when the price is real. GSA frames it as advancing federal AI adoption in line with the White House's AI Action Plan; the vendors are, plainly, buying a foothold.
That framing matters for how you read the deals. A dollar buys a trial, not a permanent entitlement. Per FedScoop, after the initial year OpenAI requires agencies to enter a paid ChatGPT Enterprise agreement or conclude access, Anthropic says it will work with the government on pricing that balances accessibility with affordability, and Google's post-2026 terms are not spelled out. The second-year cost is the number that actually shapes a program budget.
The gap the price tag hides
Here is the part the headline number obscures: access to a model is the cheap, easy part of federal AI. The hard part is everything downstream. A ChatGPT or Claude license does not by itself satisfy an Authority to Operate, keep sensitive data inside a FedRAMP-authorized boundary, produce the audit logs an inspector general will ask for, or harden a tool-calling agent against prompt injection before it touches a real system of record.
That is the work that turns a licensed model into an accredited production capability — building agentic AI systems that hold up inside ATO boundaries and audit requirements, rather than a chatbot pilot that stalls at the security review. OneGov solves procurement. It does not solve deployment, and agencies that conflate the two will find the dollar was the least expensive line in the project.
What an acquisition or program office should do
Three practical moves. First, treat the OneGov price as a runway, and model year-two costs now — for the seats you will actually keep — so the renewal is a decision, not a surprise. Second, use the trial year to learn where the model breaks against your data and your compliance posture, not just where it impresses in a demo; that knowledge is what you carry into the paid agreement.
Third, separate the license question from the system question. Deciding which model an agency can access is a procurement task GSA has largely done for you. Deciding how to deploy it safely — the boundaries, the logging, the human-in-the-loop controls — is an engineering task that starts the day the license is signed. Budget and staff for the second one, because that is where federal AI projects succeed or stall.
Sources: GSA — OneGov deal with Anthropic for Claude; FedScoop — Gemini for Government at 47 cents and the OneGov strategy. Analysis and framing by Precision AI Academy.
Common questions
What is GSA OneGov? OneGov is a General Services Administration strategy for negotiating government-wide deals with technology vendors on behalf of the whole federal government, rather than each agency negotiating separately. The AI deals put ChatGPT, Claude, and Gemini in front of agencies at nominal first-year prices.
How much do the AI tools cost agencies? About a dollar for the first year: ChatGPT Enterprise and Claude at $1 per agency, and Gemini for Government at 47 cents per agency. These are one-year introductory prices; agencies must move to a paid agreement afterward or let access lapse.
Does a OneGov license make an agency's AI compliant? No. The deals cover access to the models. Authority to Operate, FedRAMP data boundaries, audit logging, and safe agent design are separate engineering and accreditation work that the license does not address.