OpenAI’s $1-a-Year ChatGPT Deal for the Entire Federal Workforce Just Turned One. A Rival’s Protest Asked Whether It Broke the Rules. The GAO Never Answered — It Said the Challenger Couldn’t Ask.
On Aug. 6, 2025, the General Services Administration and OpenAI announced that every participating federal civilian agency could buy ChatGPT Enterprise — for any number of employees — for $1 a year. Weeks earlier, OpenAI’s own GSA schedule had listed the rate at $191 per user, per year, meaning a 1,000-employee agency’s sticker price fell from $191,000 to a single dollar. “One of the best ways to make sure AI works for everyone is to put it in the hands of the people serving the country,” OpenAI CEO Sam Altman said in the announcement.
The deal made every one of the roughly 2.29 million federal civilian employees on OPM’s books that March eligible, plus a 60-day window of unlimited access to OpenAI’s most advanced models at launch. It arrived a day after GSA cleared OpenAI, Google, and Anthropic for its Multiple Award Schedule, two weeks after the White House released its AI Action Plan, and roughly a month after OpenAI had separately picked up an unrelated $200 million Pentagon pilot contract.
Almost a year later, the discount itself is old news. What’s more interesting is what happened after: a competing AI vendor protested that the deal broke federal contracting rules, the Government Accountability Office spent four months reviewing it, and then dismissed the case without ever ruling on whether the vendor was right.
- $1 — the annual price GSA agreed to pay OpenAI, covering ChatGPT Enterprise for unlimited users at any one agency · Source: GSA, OpenAI
- $191 — the per-user, per-year rate OpenAI's GSA schedule listed just weeks before the $1 deal · Source: Government Executive
- 2.29M — federal civilian employees on OPM's rolls as of March 2025 — the baseline population eligible under the deal · Source: OPM
- Dec 19, 2025 — when GAO dismissed a rival vendor's protest against the deal — on standing grounds, without ruling on the merits · Source: GAO Docket B-423827
- Jan 9, 2026 — when ChatGPT Enterprise and OpenAI's API Platform cleared FedRAMP 20x Moderate authorization · Source: OpenAI
The mechanics of the deal were simple by government-contracting standards: one flat fee, per agency, per year, regardless of headcount. GSA Federal Acquisition Service Commissioner Josh Gruenbaum framed it as a productivity play — “improving the government’s productivity, efficiency, and ability to make better decisions on behalf of the American taxpayer” — while then-Acting Administrator Michael Rigas called it evidence GSA was “playing a leading role in the Trump Administration’s adoption of AI technology by government.” Gruenbaum later told Bloomberg that no agency would be required to renew after the first year, an important detail for a deal priced this far below cost.
The pricing didn’t exist in a vacuum. It rode in on the White House’s July 2025 AI Action Plan, overseen in part by OSTP Director Michael Kratsios, which explicitly directed agencies toward faster AI adoption. And it followed a separate June 2025 track entirely: OpenAI’s “OpenAI for Government” launch and an up-to-$200 million Defense Department pilot contract — a military procurement with its own scope, unrelated to the civilian $1 ChatGPT Enterprise agreement covered here.
OpenAI launches “OpenAI for Government” and separately wins a $200M Pentagon pilot contract — a distinct program, not the GSA deal below.
White House releases “Winning the Race: America's AI Action Plan,” directing agencies to accelerate AI adoption.
GSA adds OpenAI, Google, and Anthropic to its Multiple Award Schedule, clearing the contracting path.
GSA and OpenAI announce ChatGPT Enterprise for $1 per agency, per year — unlimited users, any agency size.
Ask Sage files GAO protests against both the OpenAI and Anthropic $1 OneGov agreements.
GAO dismisses both protests on standing grounds — it never rules on whether the deals were proper.
Edward Forst is sworn in as GSA Administrator, inheriting the OneGov AI portfolio.
ChatGPT Enterprise and OpenAI's API Platform clear FedRAMP 20x Moderate authorization.
We are providing ChatGPT access to the entire federal government for $1 per agency per year.
This is the part of the story that’s hardest to report, because the government hasn’t made it easy: neither GSA nor OpenAI has published a governmentwide dashboard showing how many of the 2.29 million eligible employees actually opened an account, at which agencies, or how often they use it. What exists instead is a scattering of agency-level coverage — FedScoop and trade outlets have tracked individual rollouts, training pushes, and a new government user community OpenAI stood up alongside the deal — but no consolidated, GSA-certified adoption number for the deal’s first year.
That opacity cuts both ways. It means claims of a runaway success story and claims of a quiet flop are both, strictly speaking, unverifiable against public data. What is verifiable: the eligible population itself has been shrinking. OPM counted 2.29 million federal civilian employees in March 2025; by February 2026 that figure had fallen to roughly 2.03 million amid broader workforce reductions — a reminder that “every federal employee” is a moving target, not a fixed base to measure adoption against.
In mid-August 2025, a competing federal AI vendor, Ask Sage, filed protests at the GAO against both the OpenAI and Anthropic $1 OneGov agreements. Ask Sage founder and CEO Nicolas Chaillan, a former Air Force chief software officer, argued the deals bypassed competition requirements, ran through reseller Carahsoft’s existing schedule contract rather than direct GSA-to-vendor terms — despite OneGov’s own stated preference for direct deals — and risked locking agencies in once they had trained staff and embedded data inside a single vendor’s tools.
“This practice not only undermines fair competition but also guarantees the squandering of billions in taxpayer dollars.”
Nicolas Chaillan, founder & CEO, Ask Sage — GAO protest filing, August 2025
On Dec. 19, 2025, GAO dismissed both protests — not by ruling the agreements proper, but by finding Ask Sage lacked standing to challenge them. Because the $1 deals were structured as modifications to Carahsoft’s existing GSA schedule contract, and Ask Sage’s own products sit on that same Carahsoft schedule, GAO treated Ask Sage as a subcontractor rather than a competing bidder. “A supplier or subcontractor for a contract is generally not an interested party to file a protest arguing that modification of the contract is outside the scope of the contract,” GAO wrote.
That is a jurisdictional finding, not a merits finding. Whether routing $1 government-wide AI deals through a reseller’s existing schedule actually complies with competition and commercial-pricing rules is a question GAO never reached — and, absent a bidder with clearer standing, may not reach again.
Whatever the legal merits, the business logic held up: within weeks, every other major AI lab had matched it. Google announced “Gemini for Government” at $0.47 per agency, per year. xAI struck an 18-month deal for Grok at $0.42 per organization — the longest OneGov AI term to date. Anthropic’s Claude agreement went further than OpenAI’s, pricing at $1 across all three branches of government, not just the executive agencies covered by the ChatGPT deal.
The rest of the year’s milestones read like a maturing bureaucracy catching up to a fast-moving vendor relationship. Edward Forst, the former Cushman & Wakefield CEO, was sworn in as GSA Administrator on Dec. 24, 2025, inheriting the OneGov AI portfolio from Rigas. Two weeks later, on Jan. 9, 2026, ChatGPT Enterprise and OpenAI’s API Platform cleared FedRAMP 20x Moderate authorization — the security baseline agencies typically require before trusting a cloud service with government data, arriving five months after the pricing deal it retroactively validated. GSA, for context, oversees more than $110 billion in products and services across federal contracts every year; the $1 AI deals are a rounding error in that ledger, but a precedent-setting one.
OpenAI is giving ChatGPT to the government for $1
One year in, three questions remain genuinely open. First, adoption: without a public dashboard, no outside party — not this outlet, not Ask Sage, not Congress — can currently verify how many federal employees actually use ChatGPT day to day. Second, legality: GAO’s dismissal resolved a standing dispute, not the underlying question of whether routing near-free AI deals through a single reseller’s schedule respects competition and pricing rules. Third, precedent: with Google, Anthropic, and xAI all having matched OpenAI’s math, the $1 model is no longer a one-off — it’s now the default way the federal government buys frontier AI, decided largely without a court or GAO ever weighing in on its legality.
A year after OpenAI cut the federal government’s ChatGPT bill by nearly 99.5 percent, the pricing has become the industry norm, not the exception — Google, Anthropic, and xAI all followed with near-identical deals. What hasn’t happened: a public adoption count, and a ruling on whether the deal was legal in the first place. GAO said the one company that asked couldn’t ask. Nobody else has.
