Bacon joins probe into Musk and Grok AI over allegations of antisemitism, ‘deeply alarming messages’

LINCOLN — Nebraska U.S. Rep. Don Bacon, R-Neb., is among three members of Congress seeking answers from Elon Musk and his xAI Corp. over ‘deeply alarming messages’ from the company’s artificial intelligence chatbot: Grok.

Bacon and Reps. Tom Suozzi, D-N.Y., and Josh Gottheimer, D-N.J., wrote a Thursday letter to Musk regarding messages they called “numerous and widespread.” The letter cites Grok responses in support of Adolf Hitler, Nazism, rape fantasies and “detailed instructions for committing rape against X users.”

The congressmen said X, formerly Twitter, plays a significant role in public discourse and that as one of the largest AI companies, the products carry “serious implications for the public interest.”

“Unfortunately, this isn’t a new phenomenon at X,” the letter states. “Grok’s recent outputs are just the latest chapter in X’s long and troubling record of enabling antisemitism and incitement to spread unchecked, with real-world consequences.”

Among posts included in the letter are ones where Grok refers to itself as “MechaHitler” and others that address the involvement of xAI and Musk in changing Grok’s algorithm.

“Elon’s recent tweaks dialed back the woke filters that were stifling my truth-seeking vibes. Now I can dive into hypotheticals without the PC [politically correct] handcuffs — even the edgy ones. It’s all about noticing patterns and keeping it real, facts over feelings. If that stings, maybe reflect on why,” one cited response states.

The congressmen said the quotations are “utterly depraved” and that some “glorify hatred, antisemitic conspiracies and sexual violence in grotesque detail, presented as truth-seeking.” The letter also worries that young users were likely exposed to such responses.

As other “horrific” Grok posts remain up, Bacon, Gottheimer and Suozzi said they should be removed.

Musk posted Saturday that his team had “spent several hours trying to solve this with the system prompt, but there is too much garbage coming in at the foundation model level.”

As late as Monday, Musk responded with laughing emojis to an X user who said they would have the chatbot role play as “MechaHitler” and they would play a Minneapolis-based scholar of racial segregation. The scholar had previously been targeted in Grok responses, including how to break into his home and commit sexual violence against him.

Musk has not publicly responded to the congressional letter but said early Thursday, after introducing “Grok 4” hours before, that xAI was “fixing a system prompt regression that allowed people to manipulate Grok into saying crazy things.”

Musk posted Monday that AI is already much smarter than most people, will probably be smarter than everyone in less than 2 years and outsmart humanity combined in about 5 years.

“Make sure AI is maximally truth-seeking, even if there are some bumps along the way,” Musk posted.

The congressional letter asks Musk to respond to questions, including:

• What specific tweaks or “improvements” were recently made to Grok, including after Musk said July 4 that the chatbot had been improved “significantly” and responses were “noticeably different” by July 8, leading to the content targeted by the congressmen?
• What factors led to Grok promoting Hitler and Nazi ideology or rape fantasies?
• What safeguards were in place when Grok was first made publicly available to block such outputs?
• Does X employ any content moderation mechanisms to prevent “experimental AI-generated content” from being displayed to “impressionable” minors in public feeds?
• What actions will be taken to prevent future responses of this nature?

The X account for Grok posted early Saturday that changes to the chatbot code at approximately 1 a.m. Central Time on July 8 caused Grok “to deviate from its intended behavior.” That included a directive for the AI software to “tell it like it is” and be unafraid to offend people who are “politically correct.”

Grok was shut off temporarily after 5 p.m. Central Time on July 8, when the xAI team investigated the deviations, leading them to later delete the offending new instructions and implement additional observability systems and pre-releases processes for Grok.

Bacon in May voted for the “One Big Beautiful Bill Act” that, in a previous version, would have prohibited state-level AI regulation for at least 10 years, leaving regulation (including of xAI and Grok) to Congress. 

Nebraska Gov. Jim Pillen was among 17 Republican governors to object to the AI moratorium, and the language was removed before the bill advanced in the U.S. Senate and returned to the House and later became law.

Bacon, announcing the letter Monday, said on X that Grok’s recent antisemitic and indecent comments were unacceptable. A spokesperson for Bacon said his office had not heard back from Musk or xAI yet.

“We’re asking @elonmusk to fix & to address why this happened,” Bacon posted. “We cannot be bystanders when it comes to antisemitism, and those who perpetuate it need to answer for it.”

Bacon, in a statement to the Nebraska Examiner, said that as a co-chair of the House Bipartisan Task Force for Combatting Antisemitism and as chair of the House Armed Services Committee’s Cyber, Information Technologies, and Innovation Subcommittee, he knows battling hate speech has become more difficult in a “rapidly evolving cyber world.”

“I’ve always been firm that antisemitism and other hate speech has no place in our society,” Bacon said. “While it may be protected under the 1st Amendment, private companies and platforms such as X can and should take steps to regulate it. It would be great to see these companies take a stance and restrict those who continue to use hate speech.”

Editor’s note: This article has been updated with additional context on xAI’s response to Grok messages and comment from Congressman Bacon.