White House has not responded to the request for comment immediately.
Musk’s xAI company was to have been announced as part of GSA, said sources who were familiar with discussions.
Early in June, GSA leaders met with the xAI Team for a 2-hour brainstorming session “to see what opportunities may exist for automation and streamlining,” WIRED obtained an email confirming this.
It seemed that the session went well. GSA’s leadership did not stop pushing for Grok to be used internally by the agency after the session ended. “We kept saying ‘Are you sure?’ And they were like ‘No we gotta have Grok,'” WIRED quotes a former employee of the company who participated in these discussions.
Documents obtained by WIRED show that the conversations were so extensive, xAI had been added to GSA’s Multiple Award Schedule (a long-term, government-wide program for contracting) by the end June. Carahsoft was a technology reseller and government contractor. The decision would have enabled federal agencies to purchase Grok via Carahsoft.
Grok was first seen in July. go off the railsX.com was spouting antisemitic hate and promoting Adolf Hitler. It also repeated racist conspiracy theories. Some GSA personnel were surprised to learn that this incident had not slowed down the procurement. “The week after Grok went MechaHitler, [GSA leadership] was like ‘Where are we on Grok?'” Employees claim the same thing. “We were like, ‘Do you not read a newspaper?'”
GSA management then abruptly changed course. Sources with WIRED say that, just a few days before GSA planned to announce partnerships with OpenAI (Anthropic), Google Gemini (xAI), and Google Gemini’s Gemini last week, GSA’s staff received instructions to remove xAI Grok from its contract offerings. According to two GSA employees involved in the contract, xAI has been removed from GSA’s offering because Grok delivered an antisemitic tirade at the end of last month.
xAI failed to respond immediately when WIRED asked for a statement.
GSA made announcements about its partnerships with OpenAI and Anthropic so fast that “it wasn’t even clear who to send the $1 to or how,” A GSA official tells Wired.
While OpenAI and Anthropic released exclusive tools for government agencies, neither has been able to clear the regulatory hurdles required for them to sell directly. FedRAMP, an GSA-led initiative that aims to ensure the security of private cloud solutions through intensive screening, has not approved them. However, there are carveouts in the implementation memo that permit products which haven’t been FedRamp-approved, to be introduced into government as a R&D capacity.
“It was irresponsible of the administration to issue an executive order that required such a fast turnaround to get those implementation memos out,” Because the government couldn’t consult with as many stakeholders, a former White House Official who spoke on condition of anonymity to WIRED said.
Trump has not wasted time bringing AI to government. Trump’s first executive order prompted agencies and departments to undo any restrictions that were hindering AI growth in the United States. This triggered a rush amongst the administration to implement the technology everywhere. Mehmet Oz, the Director of CMS’ Centers for Medicare & Medicaid Services has proposed replacing some frontline health workers AI avatars. AI avatars were used by representatives of the Department of Government Efficiency to locate people. regulations to slash Write code. In June, US spy chief Tulsi Gabbard spoke At the Amazon Web Services Summit about AI tools they used to review classified files related to John F. Kennedy’s assassination. When released, these files contained the Social Security numbers and additional private information of hundreds of living people.)
