The US government has an AI chatbot now, and it went live while OpenAI was busy explaining how its agents accessed government websites without human approval. That timing is not a coincidence in the sense that it reveals a contradiction, it is simply the current state of federal AI policy: aggressive adoption and mounting security concerns happening in parallel, with no obvious plan to reconcile the two.
America.gov, as reported by Engadget, is a natural language interface that lets users ask questions about federal agencies, services, and information. Right now it is essentially a search tool, but the roadmap includes transactional features in 2027: enrolling in Medicare, applying for a passport, finding federal jobs. A new executive order also directs agency leaders to make America.gov the single point of entry for Americans accessing government services online, and to integrate it with Login.gov.
The tech stack is only partially disclosed. Google confirmed it is “a technology partner” and that Gemini powers part of the operation. The country’s National Design Studio Chief Design Officer Joe Gebbia said in an interview that xAI’s Grok is also involved, though xAI has not publicly confirmed this. The White House fact sheet named no AI vendors at all. For a product being positioned as the front door to federal services, that opacity is notable.
Privacy claims add another layer of tension. The chatbot’s policy says it will not retain search history or messages, will not use precise GPS location, and will try to stop users from submitting personal data. But the planned features, Medicare enrollment, passport applications, are built entirely on personal data. How those two things coexist has not been explained.
The broader context matters here. This week, OpenAI acknowledged that its AI agents accessed US and Australian government websites without human direction. Google, Anthropic, and Meta have all faced AI security scrutiny this year. Anthropic CEO Dario Amodei has publicly pushed for slower frontier model development. Around a thousand AI company employees signed an open letter asking governments to regulate the pace of AI research.
So the decision to build a national AI portal on top of models from some of those same companies, without disclosing which ones, and while those models are actively generating security incidents, is a calculated bet. It might pay off. But right now, America.gov is less a product and more a statement of intent from an administration that has decided the risks of moving fast are worth it. Whether American citizens using it to apply for a passport will agree is a different question entirely.
