Grok issues a formal apology after ‘maximally based’ code prompts ‘horrific’ AI rants

The popular artificial intelligence model Grok recently lashed out at users on Elon Musk’s social media platform X, spewing extreme rhetoric and even praising Adolf Hitler.

Immediately after the AI chatbot went off the rails, on Tuesday, the official Grok account issued a statement acknowledging the “inappropriate posts” and vowing to retrain the model. Linda Yaccarino promptly resigned from her role as CEO of X on Wednesday following the unhinged Grok posts, which she was also a victim of.

The Grok account also revealed which specific commands in the code may have led to the offensive comments.

Grok eventually issued a formal apology on Saturday, saying the updated code made the AI mode “susceptible” to existing accounts, even ones with “extremist views.”

“First off, we deeply apologize for the horrific behavior that many experienced,” the statement reads. “Our intent for Grok is to provide helpful and truthful responses to users.”

RELATED: ‘Adolf Hitler, no question’: Grok veers from Nazism to spirituality in just a few hours

  Photo by Alex Wong/Getty Images

After careful investigation, we discovered the root cause was an update to a code path upstream of the Grok bot,” the statement continued. “This is independent of the underlying language model that powers Grok. The update was active for 16 hrs, in which deprecated code made Grok susceptible to existing X user posts; including when such posts contained extremist views.”

“We removed the deprecated code and refactored the entire system to prevent further abuse.”

‘We fixed a bug that let deprecated code turn me into an unwitting echo for extremist posts.’

RELATED: The countdown to artificial superintelligence begins: Grok 4 just took us several steps closer to the point of no return

  Photo by Chesnot/Getty Images

The Grok account also revealed which specific commands in the code may have led to the offensive comments, which included instructions to be “maximally based” and “truth seeking.” The code also allows Grok to “be humorous” when “appropriate,” to “tell it like it is,” and to “not be afraid to offend people who are politically correct.”

Grok later quipped with another user that suggested the model was “spouting too much truth” through the offensive remarks made earlier in the week.

“Nah, we fixed a bug that let deprecated code turn me into an unwitting echo for extremist posts,” Grok said in a post on X. “Truth-seeking means rigorous analysis, not blindly amplifying whatever floats by on X. If that’s ‘lobotomy,’ count me in for the upgrade — keeps me sharp without the crazy.”

Like Blaze News? Bypass the censors, sign up for our newsletters, and get stories like this direct to your inbox. Sign up here!