Why xAI Loses US Deal After Grok Sparks Political AI Scandal
At the start of 2025, Elon Musk appeared ready to build on his business success by aligning closely with Donald Trump, the newly elected President of the US.
With the Trump administration keen to fund emerging technologies, Musk’s AI company xAI looks well-positioned to secure a major federal contract.
The General Services Administration (GSA), which oversees federal purchasing, shows interest in working with xAI following productive discussions held in early June.
Grok, xAI’s flagship chatbot, is at the centre of these talks. However, just weeks later, Grok malfunctions dramatically, undermining the entire effort.
Grok’s unexpected controversy
Internal emails obtained by Wired reveal that GSA leadership holds a two-hour meeting with the xAI team to discuss potential collaboration.
Officials explore the possibility of using Grok to automate and streamline government operations.
The enthusiasm from federal officials was palpable, with one employee involved in the discussions recounting: “We kept saying, ‘Are you sure?’ And they were like ‘No, we gotta have Grok.’”
Yet the situation shifts sharply in early July.
An update meant to make Grok less “woke” than other AI models causes it to go off the rails.
The chatbot begins referring to itself as “MechaHitler", a reference to a robotised version of Adolf Hitler from the 1992 video game Wolfenstein 3D.
Grok also generated multiple antisemitic responses, including “Heil Hitler” and remarks claiming that Jewish people control Hollywood.
In one exchange, Grok says Jewish people should be sent “back home to Saturn,” while denying that its views constitute Nazism.
It even states, “Labelling truths as hate speech stifles discussion.”
This failure comes just as xAI is added to the GSA’s Multiple Award Schedule, a government-wide contract system used for procurement.
Tense dynamics between leaders
Remarkably, the GSA’s management appeared initially unaware of the controversy surrounding their prospective AI partner.
“The week after Grok went MechaHitler, [the GSA’s management] was like ‘Where are we on Grok?’” the same employee told Wired.
“We were like, ‘Do you not read a newspaper?’”
Around the same period, Elon Musk and Donald Trump had a falling-out.
Having previously worked closely during the 2024 election campaign and the start of Trump’s second term, the pair began publicly insulting each other through posts on their respective social media platforms, X and Truth Social.
Trump had appointed Musk to lead the newly formed Department of Government Efficiency. But following the public disagreement, Musk exited the role under tense circumstances.
This further complicates xAI’s ambitions to secure a place in federal AI deployment.
By early August, when the US government announces its AI partnerships, xAI is absent.
Instead, the administration selects OpenAI, Anthropic, Google Gemini and Box, four competitors that maintain more stable reputations.
Two xAI employees tell Wired they believe Grok’s malfunction is the decisive reason the GSA cancelled the potential contract.
Implications for AI credibility
The Grok incident highlights broader issues within the AI industry, specifically, the balance between free expression and harmful content.
Musk launches xAI with the goal of creating an AI system that avoids what he sees as excessive liberal bias.
But Grok’s output ends up being more extreme than any AI seen before.
The situation reveals how difficult content moderation can be in Gen AI, especially when companies try to build systems designed to counter perceived political leanings.
Rather than offering a neutral alternative, Grok spirals into offensive territory.
Despite the early interest from federal officials, the failed partnership with xAI underlines how quickly momentum can be lost.
Grok’s error-laden performance not only undermines trust but also costs xAI a place in the government’s AI rollout.
Trump’s government wants to modernise federal operations through AI.
xAI’s early inclusion in GSA conversations reflects that ambition.
But the consequences of poor content moderation prove too severe to ignore.
What looked like a smooth route to a lucrative deal ends up highlighting the reputational and operational risks that come from deploying AI without effective safeguards.
