Xai is about to reach a major government contract. Then Groke praised Hitler
The White House did not immediately respond to a request for comment.
A company that was supposed to be part of the GSA announcement was Musk’s XAI, according to sources familiar with the discussion.
In early June, GSA leadership held a two-hour brainstorming meeting with the XAI team to see opportunities for automation and streamlining, according to an email received by Wired.
The meeting seemed to be going well. After the end, GSA leadership continued to push the agency to launch Grok for internal use. “We kept saying ‘Are you sure?’ They were like ‘No, we have to work hard’, an employee involved in the discussion told Wired.
According to documents obtained by WIRED, the conversation went far enough that XAI was added to the GSA Multi-Reward Program, a long-term government-wide contract plan for the agency. The move will allow federal agencies to purchase Grok through Carahsoft, a technology dealer and government contractor.
Then, in early July, Groke seemed to be off track, spitting out anti-Semitic hatred, praising Adolf Hitler and imitating racist conspiracy theories on X. Some GSA employees were surprised that the incident did not seem to slow down the procurement process. “Groke went to Mechahitler a week, [GSA leadership] Like “Where are we?” the same employee claimed. “We were like, ‘Don’t you read the newspaper?’”
Then, GSA leadership seemed to suddenly change the process. Last week, GSA announced a partnership with OpenAI, Anthropic, Google’s Gemini and XAI, instructing staff to remove Xai’s Grok from contract products, two sources with knowledge. Two GSA workers involved in the contract believe that XAI was pulled due to Grok’s anti-Semitic TIRADE last month.
Xai did not immediately respond to Wired’s request for comment.
Meanwhile, GSA sources told Wired that the announcement of GSA’s partnership with OpenAI and humans was so fast that it was “not even clear who sent $1 to whom or how to send it to whom.”
Although OpenAI and Anthropic released tools tailored for government use, neither company has obtained an independent FedRamp license, allowing them to sell all products directly to the government. FedRamp is a GSA-led program that ensures the security of private cloud services through strong security screening. However, there are carvings in the implementation memorandum to allow products not approved by FedRamp to bring them to the government with research and development capabilities.
“The issuance of an executive order requires such a quick turnaround to get these implementation memorandums cancelled,” a former White House official said.
The Trump administration has not wasted time bringing AI into the administration. One of the first executive orders signed by Trump prompted institutions to reverse any rules that curb growth and dominance of American artificial intelligence, launching crazy actions among government leaders in search of new ways to incorporate technology into everything. At Medicare and Medicaid Service Centers, Mehmet Oz recommends using AI avatars to replace some frontline health workers. Representatives of the so-called government efficiency ministry use AI to find regulations to cut and write code. In June, U.S. spy chief Tulsi Gabbard delivered a speech at the Amazon Web Services Summit on using AI tools to review confidential documents related to the assassination of John F. Kennedy. (After release, the documents turned out to contain the social security numbers and other private information of hundreds of living people.)
