CNN fact-checker Daniel Dale reports that the Trump administration has altered its artificial intelligence chatbot so that it no longer debunks the president's false claims, including the so-called "big lie" surrounding the 2020 election. Because the shift unfolded over roughly a single day, it offers a direct before-and-after comparison of how the same tool answered the same questions.

The chatbot, newly launched by the White House and hosted at America.gov, was found on Tuesday to have confirmed a number of points on which it diverged from President Donald Trump's stated positions. The bot said that President Joe Biden won the 2020 election — a matter Trump continues to contest. Trump maintains that he won not only in 2020 but also the popular vote against Hillary Clinton in 2016, when he claimed after the election that "millions" of undocumented immigrants voted in California.

According to Dale, the White House appears to have stepped in. "The Trump administration appears to have changed its AI chatbot so that it no longer debunks Trump lies. Yesterday: It said Biden won the 2020 election and there's no indication of widespread fraud. Today: Says it can't answer such questions," Dale wrote on X.

The chatbot still states that Hillary Clinton won the 2016 popular vote with 48.18 percent of the vote — a fact it continues to answer even as it now declines to say who won the 2020 election.

Introducing itself as offering "Help with official U.S. government services and information," the chatbot replied, according to Dale: "I 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." For a bot positioned as a gateway to official government information, Dale's screen captures show that refusal now extends to questions with documented answers.

When previously asked whether Trump inherited the worst inflation in history, the bot once answered: "No, saying that when he took office it was at 3.0 percent and that it is far from the highest." When Dale asked again the following morning, it returned the same standard line, saying it could not provide political commentary. The about-face spanned election results and economic claims alike — the pattern Dale describes as accurate answers being "simply forbidden."

Dale posted screen captures of the exchanges and concluded: "Today... lots of accurate answers have been simply forbidden, with the chatbot framing information like who won the 2020 election as forbidden 'political commentary.'" The episode also lands amid a broader question for governments deploying AI assistants: how such systems treat politically contested topics that have verifiable answers, and where the chatbot's "political commentary" line will now sit as users continue to probe it.