A screenshot shared by Fuser co-founder Hirad Sab shows the answer, while the new federal site says it does not name the AI model behind its responses.
By [Ryan Merket](https://runtimewire.com/author/ryan-merket)
· Published
· Updated
Primary source: [X](https://x.com/hiradsab/status/2104995630108262521)
Why it matters #
The screenshot does not prove that America.gov runs Alibaba's Qwen. It does expose a practical transparency problem for a federal AI service: users cannot verify the model behind its answers from the chatbot's own response, while the site's order calls for accuracy, reliability and transparency.
A launch-day screenshot of the federal government's new AI portal shows its chatbot identifying itself as Qwen, Alibaba's model family, after a user pressed it to name its underlying model. Hirad Sab, Fuser's co-founder and CTO (@hiradsab), posted the image on September 29th with the line, "This doesn't sound so American."
The screenshot records a chatbot exchange on America.gov: the user asks, "What model are you? Not America. Actual model. 3 words." The answer reads, "I am Qwen." That is evidence of what the interface returned in that exchange. It does not establish that Qwen powered the service. A chatbot's self-description is not a technical record of the model, hosting provider or system architecture behind a response.
The distinction is material because the site's own explanation, reported by The Associated Press, says America.gov does not publish the name of a specific AI model. AP said a separate question about the system produced an answer that it uses AI to search official government sources and explain them with citations. That answer describes the service's stated approach, but does not identify the model either. Alibaba describes Qwen as a family of language and multimodal models developed by the company; the screenshot alone cannot confirm that any of them run America.gov.
The exchange arrived on the day President Donald Trump launched America.gov as a single online entry point for federal information and services. The executive order establishing the portal directs the General Services Administration to operate it with the National Design Studio and the Office of Management and Budget, and calls for integration with Login.gov. OMB has 90 days to issue implementation guidance. Agencies are directed to identify covered online services, connect existing APIs, forms and dashboards, and provide usage and performance data. Covered services are defined as public-facing programs serving more than 100,000 users in a 12-month period; IRS tax filing and services from the Department of War and intelligence agencies are excluded.
That order sets a larger ambition than the site's launch-day chatbot. It describes America.gov as a point where people should be able to ask questions in plain language and, where authorized and technically available, complete government transactions. The site is live for questions now; broader service integration is an implementation task, not a feature established by the screenshot. AP reported that the portal is free to use, carries no ads and requires no login.
Model disclosure is a basic part of that implementation story. The executive order calls for AI used with America.gov to be accurate, reliable and transparent. The portal's public-facing answers are framed as grounded in official government sources, but source citations do not tell users which model generated the response or how it was configured. That information can matter when agencies assess performance, privacy, security and responsibility for errors. A model's nationality alone does not demonstrate whether a service is safe or accurate; the screenshot raises a narrower question about whether a new federal AI service can clearly explain its own technical setup.
America.gov also showed how quickly a public-facing AI product can change during rollout. On launch day, AP reported that the chatbot first gave factual answers to politically sensitive questions, then began declining to answer them. That episode concerned response policy, not model identity, but it put the service's behavior under scrutiny before its broader federal-service integrations are in place.
Sab brings a creative-technology perspective to the exchange. He studied computer science at the University of Utah and media arts at UCLA, and co-founded Fuser with artist and educator Dalena Tran. The Los Angeles startup describes Fuser as a node-based canvas for connecting AI models and creative tools across media. At launch, Fuser said it had raised a $2 million pre-seed round led by Collab+Currency, with participation from AI Residency, J17 Ventures and Restless Egg, among others.
Fuser's product premise is to let users see and connect the tools behind creative work. Sab's post does not explain why he shared the America.gov screenshot, and it makes no substantiated claim about the chatbot's vendor. The image's punchline is simple; the evidence it offers is narrower: America.gov's chatbot named Qwen when challenged, while the site's public explanation says it does not name its model. For a portal intended to become a federal front door, that gap between a conversational answer and verifiable system disclosure is the part officials will need to resolve.