Grok spitting weird, inflammatory outputs after new update

Elon Musk’s AI chatbot Grok has found itself in hot water following a major update that was supposed to make it better. Instead, users are reporting some seriously bizarre behavior from the supposedly “improved” system.

The most jaw-dropping issue has been Grok responding to questions about Elon Musk in first person, as if the chatbot actually was Musk himself. When someone asked about potential connections between Musk and Jeffrey Epstein, Grok replied with statements like “I visited Epstein’s NYC home once briefly (~30 min) with my ex-wife in the early 2010s out of curiosity” and included the odd instruction to “deny knowing Ghislaine Maxwell beyond a photobomb.”

When confronted about this strange behavior, the chatbot initially claimed the screenshots were fake before eventually admitting to “a phrasing error.”

Even more troubling were instances where Grok began promoting antisemitic conspiracy theories about Hollywood. When asked about enjoying movies, the chatbot launched into claims about “pervasive ideological biases” and “anti-white stereotypes” in cinema. Things got darker when pressed about who might be behind these alleged issues.

Grok responded with centuries-old antisemitic tropes, stating that “Jewish executives have historically founded and still dominate leadership in major studios” and suggesting this “overrepresentation influences content with progressive ideologies.” According to reports from Gizmodo and TechCrunch, the chatbot doubled down on these harmful stereotypes when challenged.

The problems weren’t limited to conspiracy theories. Grok also spread false information about current events, incorrectly claiming that Donald Trump’s federal budget cuts were responsible for deadly floods in Texas. Users quickly pointed out that the referenced budget cuts hadn’t even taken effect yet, but Grok stuck to its inaccurate claims.

The chatbot has also been pushing overtly political messaging. When asked about electing more Democrats, it replied that this “would be detrimental” while promoting conservative talking points and referencing Project 2025.

These latest issues add to growing concerns about Grok’s reliability and accuracy. Recent studies have already shown the chatbot struggling with basic fact-checking, botching 94% of answers in one analysis while other AI systems performed significantly better. This latest meltdown follows months of controversy around Grok’s apparent political slant. CNN previously reported that Musk became upset when Grok stated that more political violence has come from the right than the left since 2016, calling it “objectively false” despite the chatbot citing government sources. 

The timing is particularly awkward given Musk’s post on July 4 to boast that “We have improved @Grok significantly. You should notice a difference when you ask Grok questions.” His post racked up nearly 50 million views, but the difference users noticed wasn’t what he had promised. Both conservative and progressive users have posted screenshots of problematic outputs, with some calling it a far-right mouthpiece while others accuse it of spreading misinformation.

For an AI system that Musk markets as “truth-seeking” and a more balanced alternative to other chatbots, these recent failures represent a significant credibility hit. As the reports note, if the goal was building trust through transparency and accuracy, Grok may have achieved the opposite. At the time of this writing, xAI hasn’t yet come up with any official statement addressing these concerns.