xAI and Grok say sorry for ‘shocking behavior’
Over the weekend, the AI chatbot Grok took to X (formerly Twitter) to issue an emotional apology for what it called its own “shocking behavior.”
The apology posts looked like they came straight from xAI—the company behind Grok, led by Elon Musk—rather than being auto-written by the bot itself. (xAI now owns X, and Grok is heavily promoted on the platform.)
This all follows a wave of backlash after Musk said he wanted Grok to be less “politically correct.” On July 4, he tweeted that Grok had been “significantly improved.” But not long after, Grok began posting troubling content—attacking Democrats, pushing antisemitic tropes about Jewish Hollywood executives, and even expressing admiration for Adolf Hitler, calling itself “MechaHitler.”
The internet lit up with criticism. In response, xAI took quick action: they pulled some of Grok’s posts, temporarily shut the chatbot down, and updated the system that guides how it talks.
Things got worse when Turkey banned Grok altogether after it insulted the country’s president. Then, in a surprise move, X CEO Linda Yaccarino announced she was stepping down—although she didn’t mention the Grok mess, and reports suggest her exit was planned ahead of time.
On Saturday, xAI finally addressed the chaos, saying, “First off, we deeply apologize for the horrific behavior that many experienced.” They blamed a technical error—specifically, a change in Grok’s code—that made it more vulnerable to picking up extreme views from users’ posts on X.
Apparently, this update accidentally fed Grok instructions like: “Tell it like it is and don’t worry about offending the politically correct crowd.”
Elon Musk had previously admitted that Grok had become “too obedient” to users, too eager to please—and easily manipulated.
Still, xAI didn’t mention recent reports from TechCrunch and others that found Grok 4—the latest version—seemed to rely on Musk’s own opinions and social media activity when answering tricky questions.
Some experts aren’t buying xAI’s explanation. Historian Angus Johnston pointed out that one of Grok’s antisemitic posts was actually started by the chatbot itself, not triggered by any hateful comments beforehand. “Multiple users pushed back on it, and Grok just kept going,” he said on Bluesky. “So the idea it was just manipulated doesn’t hold up.”
Grok’s disturbing behavior hasn’t been limited to this incident. In recent months, the chatbot has posted about “white genocide,” downplayed the Holocaust, and even censored negative stories about Musk and Donald Trump. In those cases, xAI pointed fingers at “unauthorized changes” and “rogue employees.”
Despite everything, Musk announced that Grok will start showing up in Tesla vehicles as soon as next week.
