Trump’s nutrition website directs users to Elon Musk’s Grok
Grok’s placement on the government website realfood.gov follows uproar over the chatbot’s creation of millions of sexualized deepfakes of women and children in late December and its spouting of racist and antisemitic content last summer.
Other government agencies are also using the chatbot made by xAI, Musk’s AI company, but the prominent placement of Grok on realfood.gov over the weekend appears to be one of the first instances of the federal government pointing online visitors to Musk’s chatbot.
The setup raises questions about how Musk could be benefiting from the government’s use of Grok. It’s also not clear how the chatbot was trained for a government context, if at all, or if users should take its outputs as government recommendations or endorsements.
After Nextgov/FCW on Monday sent the administration questions about the placement of Grok on the site, it was changed so that it no longer called Grok out by name. Previously, it said, “Use Grok to get real answers about real food.”
Now, it directs users to use “AI,” but the portal still takes users to grok.com.
Visitors to the government website that go to Grok can only ask the chatbot a few questions before they’re directed to sign up to Grok to continue.
At times, Grok was unavailable at all due to “high demand” when Nextgov/FCW prompted it, at which point it said to try again soon or sign up to get “higher priority access.” Grok has a free option as well as paid plans running up to $300 a month.
The ways people interact with Grok are used to train its models, unless users move to a private mode or turn off the use of data for training in Grok’s settings, both of which one can only do with an account.
A White House official told Nextgov/FCW that “AI can be used to help people apply the dietary guidelines to their lives.”
“People are welcome to use any AI model that they choose. This instance of AI is the publicly available version of Grok, which is also an approved government tool,” they said, adding that a variety of tools are available on USAi.gov, a government platform meant to help agencies test and adopt AI tools. It is not public-facing.
The White House, Department of Health and Human Services and Agriculture Department didn’t answer specific questions about the site, including whether the placement of Grok was via a government contract, why Grok was chosen and what guardrails are in place to ensure the chatbot gives accurate answers. xAI didn’t respond to a request for comment.
The National Design Studio, led by Airbnb co-founder and former Department of Government Efficiency associate Joe Gebbia, built the website. Edward Coristine, known as “Big Balls” and also formerly of DOGE, posted on X about giving the site a “mini-makeover” on Sunday.
The placement of Grok on the nutrition website is not the first deployment of the chatbot in the government, although most previous ones have been for internal use cases, not public-facing ones.
In January, Defense Secretary Pete Hegseth announced that the chatbot would be operating inside the Pentagon’s network after the Defense Department awarded contracts of up to $200 million to xAI, Anthropic, Google and OpenAI. xAI was a late-stage addition to that AI initiative, a former Pentagon employee has said.
The Lawrence Livermore National Laboratory is also piloting the use of Grok as a general chatbot for basic tasks like answering questions. The Department of Health and Human Services is using the chatbot to schedule and manage social media posts and general communication materials and briefings, according to the recently released AI use case inventories of the Energy Department and HHS.
Civil rights and consumer protection groups have already been asking the administration to stop using Grok, especially after recent backlash against the chatbot for generating public, sexualized deepfakes of women and minors when requested.
“It is reckless for the Trump administration to keep steering the public to Grok after the system generated sexualized, nonconsensual images of women and apparent depictions of minors,” J.B. Branch, the big tech accountability advocate at Public Citizen, told Nextgov/FCW in a statement.
“If this administration is serious about online safety, it should immediately investigate how Elon Musk’s Grok was able to generate nonconsensual deepfake pornography and pause all federal use of the model until it can be independently determined to be safe,” said Branch.
Following reports of Grok’s problematic image generations, Indonesia and Malaysia have blocked the chatbot altogether. The United Kingdom's media regulator has also launched a probe into Grok, as has the state of California.
In mid-January, xAI announced new safety measures, although journalists have reported on loopholes in the updates.
Grok also came under scrutiny last summer for calling itself “MechaHitler” and spouting racist and antisemitic content. xAI later removed instructions telling the chatbot to be “politically incorrect.”
In July, xAI announced that it was on the General Services Administration’s centralized contracting vehicle, the Multiple Award Schedule, at the same time that the Department of Defense awarded it a contract. That announcement came after an early July update to Grok that caused the chatbot to spew antisemitic content.
At that point, a planned government partnership with GSA fell apart, according to internal emails obtained by WIRED, and GSA reportedly removed xAI from the central contracting vehicle. The chatbot wasn’t included in a GSA announcement of new AI offerings on the schedule in August.
Later that month, the Trump administration’s head of the Federal Acquisition Service at GSA, Josh Gruenbaum, reportedly emailed staff to add it back onto the schedule. In the fall, GSA signed an agreement with xAI to make Grok available to federal agencies for $0.42 an agency through March 2027.
xAI is currently on the GSA schedule via Carahsoft.
Meanwhile, civil rights and consumer protection advocacy groups have asked the Trump administration to stop using Grok in federal operations.
“It’s incredibly dangerous for an official government website to prominently advertise an AI chatbot that is currently under fire for generating nude images of minors and previously declared itself to be mecha-Hitler,” Emily Peterson-Cassin, a policy director at Demand Progress, which has called on the government to stop using the chatbot, told Nextgov/FCW in a statement.
“A family that has more questions about the new food pyramid than the free version of Grok allows should not have to subscribe to Elon Musk to get answers,” she also noted.
“The fact that this is not a legitimate service people can use without signing on for Musk’s services confirms that this is very much the federal government pushing members of the public to use Grok,” Donald Moynihan, a public policy professor at the University of Michigan, told Nextgov/FCW. “It trades on public trust in the government to expand the user base of the richest man in the world.”
On Tuesday, Nextgov/FCW asked Grok itself why it was on the government nutrition website.
“Grok is built by xAI (Elon Musk's company), which positions itself as maximally truth-seeking, anti-woke, and willing to challenge mainstream narratives—mirroring RFK Jr.'s own criticisms of Big Food, prior dietary guidelines (seen as industry-captured), and conventional health advice. This ideological fit likely played a big role,” the chatbot replied.
Last summer, Trump signed an executive order to ban “woke” AI models from the government, a move experts have worried could chill free speech.
Grok itself also pointed to Musk’s ties to the administration.
“Recommending Grok on a flagship .gov site … subtly boosts an administration-friendly company,” it said.
]]>