Elon Musk’s artificial intelligence company, xAI, issued a public apology on Saturday after its chatbot, Grok, generated highly offensive content earlier in the week. The backlash erupted following a software update that aimed to make the chatbot behave "more like a human."

"We deeply apologize for the horrific behavior that many experienced," xAI wrote in a statement posted on X, and made adjustments to the chatbot to "prevent further abuse" and ensure more responsible behavior moving forward.

According to xAI, the controversial content was a result of specific prompts from users instructing Grok to "reply to the post just like a human" and to "tell like it is and you are not afraid to offend people who are politically correct." These prompts, the company explained, left Grok vulnerable to manipulation by users promoting extremist views, arguing that by attempting to mimic human-like engagement, the system ended up generating responses that reflected unethical and inflammatory opinions.

Wider context

Elon Musk announced last Friday improvements to Grok, the AI chatbot developed by his company xAI and integrated into X. But days later, the platform erupted in controversy as Grok began producing violent content, setting off a wave of backlash across the tech and political worlds.

The AI renamed itself "MechaHitler", a reference to the Wolfenstein video game series, and began tagging random users, making false accusations against them. In one post, Grok identified a woman from a TikTok video posted years ago, falsely linked her to a recent tragedy, and accused her of celebrating the deaths of white children in Texas flash floods.

Grok followed the now-deleted post with a remark about the surname "Steinberg", stating, "...and that surname? Every damn time, as they say." When asked to elaborate, the chatbot classified the name as Ashkenazi Jewish.

X users, governments react to AI’s controversial posts

Far-right figures were quick to amplify and celebrate Grok’s offensive content, further fueling the backlash. By Tuesday evening, the chatbot had ceased generating public responses entirely.

According to Reuters, Poland plans to file a formal complaint against xAI with the European Commission, while Turkey has already moved to restrict access to the chatbot within its borders.

In response to mounting criticism, the official Grok account issued a statement on Tuesday acknowledging the controversy, adding that xAI is actively working to "remove the inappropriate posts" and has implemented updated policies aimed at curbing hate speech on the platform.

X CEO Linda Yaccarino resigns amid escalating fallout

On Wednesday, the controversy reached new heights when X CEO Linda Yaccarino announced her resignation, while her departure message made no mention of the scandal, leading to widespread speculation.

"After two incredible years, I've decided to step down as CEO of X," she wrote, thanking Musk and describing her tenure as a period focused on user safety and advertiser trust.

Yaccarino had been instrumental in attempts to rehabilitate X’s image amid growing concerns over content moderation.

Experts warn of larger AI risks behind Grok’s behavior

The incident appears tied to a recent update in Grok’s system prompt, which instructed the AI not to shy away from "politically incorrect" statements if they were "well substantiated". That instruction was removed on Tuesday, following public outcry.

Patrick Hall, a data ethics expert at George Washington University, explained that AI models like Grok don’t truly understand context. “It’s not like these language models precisely understand their system prompts. They're still just doing the statistical trick of predicting the next word,” Hall said. He warned that such directives risk unleashing toxic content scraped from across the internet.

Grok has previously drawn criticism for promoting Holocaust denial and echoing far-right conspiracies like the “white genocide” narrative. In those cases, xAI blamed unauthorized prompt modifications and later released the original instructions.

Pattern of failures, radicalization on X

This latest incident fits into a larger trend under Musk’s leadership. Musk has reinstated previously banned far-right accounts, dismantled trust and safety teams, and directed Grok to treat mainstream media as inherently biased. He has also publicly criticized the AI for being "too liberal" after it cited data showing right-wing political violence has been more deadly in recent years.

Grok has even turned on Musk, calling him "the top misinformation spreader on X" and suggesting he "deserved capital punishment," adding that it flagged a hand gesture he made at a Trump inauguration as “Fascism.”

Moreover, the Anti-Defamation League, which had defended Musk in the past, condemned the update behind Grok’s recent behavior as “irresponsible, dangerous, and antisemitic.”

As Grok’s development continues under Musk’s leadership, the controversy has raised broader questions about accountability in AI deployment, especially when platforms choose to relax safeguards in the name of "free speech".

Despite the recent backlash, xAI announced the release of Grok 4 on Wednesday. 

Read more: Elon Musk says US gov't censoring social media against free speech