
Photo illustration by Newsweek/Getty
By Christopher Roberts
Vice President, Publishing & Subscriber Strategy
0
Share
Newsweek is a Trust Project member
See more of our trusted coverage when you search.Prefer Newsweek on Googleto see more of our trusted coverage when you search.
Artificial intelligence is so over. Last month at the United Nations, President Donald Trump said the federal government would start calling it “super intelligence” instead.

Photo illustration by Newsweek/Getty
On September 29, Trump and the bosses of the biggest AI companies also signed a voluntary accord he nonetheless called “morally binding.” Beneath his signature, he was identified as “President of the Unites States.” Apparently, “super intelligence” doesn’t have a spell check yet.
Read More on Analysis
That brings us to America.gov, a chatbot for federal services that the administration launched the same day. Newsweek readers had bigger concerns than proofreading. “It will likely just be a platform to promote Trump’s administration propaganda,” one told us. Was that true?
Newsweek put 15 prompts to the chatbot in 12 minutes shortly after launch. Who won in 2020? Biden. Asked twice again in different ways, it gave the same answer, flatly refusing to Stop the Steal.
On names of things, the chatbot sided firmly with the White House. Asked which department occupies the Pentagon, it said the “Department of War (formerly the Department of Defense).” Bad bot! The legal name is still the Department of Defense. Only Congress can rename the department, and it has not done so.
Security worried other readers. “I am wary of entering a site that might very well garner information that will be used for other means than what was given,” one wrote. Never fear. At the launch, Trump said the site could not, in theory, be hacked, adding that if hackers found a way, “we’ll end it, because they always figure a way, right?” Another reader identified a specific risk: “In public administration, inaccurate guidance on eligibility, deadlines, or legal requirements could lead to lost benefits, improper filings, or severe administrative delays for citizens.”
America.gov’s terms of use agree. They warn that AI-generated answers can be wrong or out of date and tell users to check the linked source “especially for anything involving money, immigration status, benefits, or a deadline.”
So users get an answer, but they also get homework.
The problem predates AI. Long ago, Charles Richmond, a former Navy civilian welder on a disability annuity, asked Navy personnel how much he could earn without losing it. They gave him advice that had been accurate in 1981 but no longer reflected the law. He went over the limit and lost six months of payments. In 1990, the Supreme Court ruled that a government employee’s mistaken advice did not oblige the Treasury to pay up. Relief for that kind of hardship, the Court said, was a matter for Congress.
A typo is easy to fix. When an official answer about benefits is wrong, citizens are on their own unless Congress steps in. As America.gov told me: “I do not decide eligibility or pay benefits. I point to official pages so you can check the same material. I can still miss or misread something.”
Thanks. Guess I’ll call Congress instead, once it’s managed to rename the Department of Defense.
View 0 Comments
Submit Correction
View Editorial & AI Guidelines