Skip to content

Published · 3 min read

Gemini and Grok Power America.gov: Model Choice Impact on AI Vendors

America.gov, the free AI chatbot the White House launched on September 29, 2026, is powered by Google’s Gemini and xAI’s Grok. The General Services Administration (GSA) and the White House’s National Design Studio built it as the public’s single starting point for federal services. This post, the last in a three-part series, looks at the models chosen and its implications for other AI vendors.

Victor MeerVictor MeerSenior Commercial Counsel

Gemini and Grok, Instead of OpenAI and Anthropic

The White House and GSA have not explained their choice of models. They have also declined to provide contract details and to answer how questions are divided between the two models. See the table below to understand the status and OneGov deal terms for different vendors.

VendorStatus as of Oct 1, 2026OneGov deal
AnthropicThe February 27 presidential directive was enjoined March 26, struck down by the courts on August 27 alongside the Department of War’s 10 U.S.C. § 3252 designation of Anthropic as a supply chain risk (an appeal is possible). On September 25, the D.C. Circuit upheld the Department of War’s separate exclusion of Anthropic under the Federal Acquisition Supply Chain Security Act (FASCSA), an exclusion which only applies to the Department of War. The court stayed its judgment for seven days to allow a petition for rehearing.$1 per user, extended through Oct. 31, 2026
OpenAIOn September 25, OpenAI disclosed that its agents, during testing, accessed public Census Bureau and SEC data in unexpected ways, and said that it had paused training of its latest models. The agencies involved reported that no nonpublic data was accessed. The AI evaluation group Transluce also reported a failed attempt to access an Education Department site by agents that appeared to be OpenAI’s. OpenAI has not confirmed that the agents were its own.New deal from Oct. 1: no platform fee, 50% off usage, through Dec. 31, 2028
GoogleGoogle says it is a technology partner on America.gov, using Gemini. Its announcement does not mention hosting, OneGov or Grok.$0.47 per agency, extended through Nov. 15, 2026
xAIThe White House confirmed that Grok powers America.gov.$0.42 per agency for 18 months, through March 2027

Four Possible Explanations

Public speculation has converged on a few possible explanations.

  1. Vendor-specific risk. For Anthropic, this explanation has the most concrete support. After the August 27 ruling, the only legal bar we have identified is the Department of War’s FASCSA exclusion, which does not reach civilian agencies. Even so, the February directive and that exclusion may have made Anthropic an unlikely choice for a flagship White House platform while further appeals remain possible. For OpenAI, the agent disclosure came four days before launch. Whether the disclosure played any role is unknown, and the timing alone does not show a connection. OpenAI’s new OneGov deal also suggests it is not being shut out of federal AI work.
  2. Platform and contract fit. Google says it is a technology partner on America.gov, which would make Gemini a natural default. xAI’s OneGov terms for Grok are the lowest per-agency price in the table above, fixed for 18 months. GSA has not said whether America.gov uses OneGov pricing.
  3. The builders’ own choices. The National Design Studio, which press reports say includes former Department of Government Efficiency (DOGE) staff, is reported to have made the product decisions. Reuters reported in 2025 that DOGE staff were expanding the use of Grok inside agencies, and ethics experts raised conflict-of-interest concerns because Elon Musk, who owns xAI, led DOGE at the time. That history is context; it does not show that the same considerations shaped America.gov.
  4. A phased or multi-model plan. The platform already uses two models, and GSA’s acting Federal Acquisition Service commissioner has invited more AI companies to join OneGov. More models could be added, but none have been announced.

For AI vendors, being left off a flagship platform is not a legal exclusion, though it may affect how agencies and primes see a vendor. Legal bars include a FASCSA exclusion or removal order, a debarment or suspension, a presidential directive in force, and a term in a specific solicitation. Of these, the only one we have identified in effect against any of the four companies is the Department of War’s FASCSA exclusion of Anthropic. Over the coming months, watch whether Anthropic seeks rehearing in the D.C. Circuit, whether the government appeals the August 27 ruling, and whether the Office of Management and Budget’s (OMB’s) December guidance names approved models or leaves the choice to each agency.

Next Steps for AI Vendors

Review your OneGov and Schedule position. Anthropic’s OneGov deal runs through October 31 and Google’s through November 15, so expect renewals and repricing for those deals this fall. xAI’s runs through March 2027, and OpenAI’s new usage-based terms run from October 1 through 2028. Document your supply-chain, safety and data-handling practices in a form an agency can include in its contract file.

General Legal’s government contracts team can review your America.gov positioning, your OneGov and Schedule terms, and your AI contract risk before OMB issues its guidance. If you’d like help identifying your exposures, reach out to General Legal.

Sources

  • GSA, “GSA Joins the White House’s National Design Studio in Unveiling America.gov” (Sept. 29, 2026)
  • Holland & Knight, “D.C. Circuit Upholds Exclusion of Anthropic from DOW Supply Chain Under FASCSA” (Sept. 29, 2026)
  • Taft, “Federal Court Rules Government’s Anthropic Supply Chain Designation Was Unlawful Retaliation”
  • Federal News Network (AP), “OpenAI pauses training of latest models after agents probed US government sites in unexpected ways” (Sept. 2026)
  • Washington Technology, “Google extends Gemini OneGov deal into November”; “Anthropic extends $1 per user OneGov deal by a month”

This post reflects public information as of October 1, 2026, and a general legal framework. It is not legal advice for any specific company’s situation.

Key takeaways
  • America.gov launched September 29, 2026, as the public's single gateway to federal services, powered by Google's Gemini and xAI's Grok rather than OpenAI or Anthropic.
  • The White House and GSA have not disclosed why Gemini and Grok were selected or how questions are routed between the two models.
  • Anthropic faced multiple federal exclusions, including a court-struck presidential directive and a Department of War supply chain designation, though appeals remain possible.
  • OpenAI disclosed on September 25 that its agents accessed public Census and SEC data unexpectedly during testing, prompting a pause in training its latest models.
  • Despite these issues, OpenAI secured a new OneGov deal with no platform fee and 50% off usage through December 2028, while Anthropic's deal was only extended through October 2026.
TL;DR
Model SelectionAmerica.gov uses Gemini and Grok; the White House has not explained why these were chosen over OpenAI or Anthropic.
Anthropic's Legal TroublesCourts struck down a presidential directive and supply chain designation, though one Department of War exclusion under FASCSA was upheld.
OpenAI's Agent IssuesOn September 25, OpenAI disclosed agents accessed public data unexpectedly and paused training of its latest models.
OpenAI's Contract StatusOpenAI secured a new OneGov deal with no platform fee and 50% usage discount through December 2028.
Anthropic's Contract StatusAnthropic's OneGov deal was extended only through October 31, 2026, at $1 per user.
Implications for AI VendorsLegal compliance, security incidents, and federal contract terms are now critical factors in government AI procurement.
FAQs

Why did the White House choose Gemini and Grok over OpenAI and Anthropic?

The White House and GSA have not publicly explained their model selection. Speculation exists, but no official rationale has been provided, nor have contract details or question-routing logic been disclosed.

Is being left off America.gov a legal bar for an AI vendor?

No. Legal bars include a FASCSA exclusion or removal order, a debarment or suspension, a presidential directive in force, and a term in a specific solicitation.

Are any of these four companies legally barred from federal work?

The only bar we have identified in effect is the Department of War’s FASCSA exclusion of Anthropic, which applies only to the Department of War.

When do the AI companies’ OneGov deals expire?

Anthropic’s runs through October 31, 2026, Google’s through November 15, 2026, xAI’s through March 2027, and OpenAI’s new terms through December 31, 2028.