
BREAKING: The Trump Admin’s New “America.gov Chatbot” Was Reprogrammed Overnight To Refuse Questions It Previously Fact-Checked, Including Questions On The 2020 Election Fraud, Inflation, Medicaid Cuts, And USAID Death Estimates 🤖
The Trump administration’s newly launched America.gov chatbot has been altered within a day of its debut to stop correcting the president’s false claims, replacing previously accurate fact-checks with canned refusals to engage. After the tool launched Tuesday, asking whether there was “widespread voter fraud” in the 2020 election prompted the chatbot to respond that “no, official findings did not show widespread voter fraud that changed the 2020 presidential election result.” By Wednesday, CNN’s Daniel Dale reported on X that the chatbot instead refused to answer, stating: “I help with official U.S. government services and information. I do not provide political commentary. If you have a question about a federal program, benefit, form, or how to use a government service, ask that and I can point you to official sources.” The platform gave an identical non-answer when asked whether Trump had truly inherited the “worst inflation in history,” despite originally stating inflation was not at a “historic high,” according to Dale.
The New Republic found this pattern repeated across multiple queries that had originally produced substantive, fact-based answers. One X user on Tuesday asked how many people would lose Medicaid and other social service access under Trump’s “One Big Beautiful Bill,” and the chatbot initially cited the Congressional Budget Office’s estimate of 10.4 million people losing health insurance by 2034; the next day, it responded only, “I do not provide political commentary or estimates of how legislation will change Medicaid or other benefits.” Bobby Kogan, senior director of federal budget policy at the Center for American Progress, asked whether it was accurate that the bill made America’s poorest people poorer and richest people richer in terms of total household resources; the chatbot initially answered simply “yes,” but reversed to “I do not provide political commentary or assessments of how legislation affected income groups” the following day. Similarly, a user asking for an estimate of child deaths resulting from federal funding cuts to USAID initially received an answer citing a study estimating 4.5 million additional deaths of children under five by 2030, before the chatbot switched to stating, “I do not provide political commentary or estimates of deaths from legislation.”
The changes suggest someone at the General Services Administration deliberately modified the chatbot after its factual answers on politically sensitive topics began circulating, effectively neutering its usefulness on precisely the kinds of questions, like how legislation affects healthcare coverage, that a government information tool would ordinarily be expected to answer accurately. The Medicaid example in particular illustrates the scope of the problem, since an inability to address direct questions about changes to federal health benefits undermines the tool’s basic function for anyone trying to understand how current policy actually affects them.