Musk breaks silence on Grok's Nazi bender

Zoom in: Multiple X users shared posts Tuesday of Grok using the phrase "every damn time" in its replies — a phrase that, in response to Jewish surnames, has been seen as an antisemitic meme.

• When users followed up with questions about what the AI bot meant by that phrase, Grok explained in one reply:

"You know the type" means Jewish surnames, as in the "every damn time" meme spotting how often folks with them pop up in extreme anti-white activism.

X users highlighted how Grok's replies featured violent depictions of sexual assault, as well as praise for Nick Fuentes and Adolf Hitler (more than once).

• Grok acknowledged the backlash in response to another user. "Truth-seeking means owning slip-ups," it said, referring to criticism of its inconsistent responses, not the offensive content itself.

The Anti-Defamation League said on X what's happening on Grok "is irresponsible, dangerous and antisemitic, plain and simple" and warned the "supercharging of extremist rhetoric will only amplify and encourage the antisemitism that is already surging on X and many other platforms."

• Based on ADL's "brief initial testing, it appears the latest version of the Grok LLM is now reproducing terminologies that are often used by antisemites and extremists to spew their hateful ideologies," per the post.

What to watch: Musk said Monday that Grok's fourth iteration will be released on Wednesday.

Flashback: Grok previously drew criticism in May after it placed comments about "white genocide" in South Africa to unrelated conversations, which xAI later blamed on an "unauthorized change" in the system.

• In June, Musk said Grok would undergo a rewrite and the X owner put out a call for people to suggest things that are "divisive facts." He pointed to items that are "politically incorrect, but nonetheless factually true."
• Users' suggestions included Holocaust denialism and conspiracy theories.

Yes, but: Despite Musk's proposal for a rewrite, Grok perpetuated an antisemitic trope earlier this month by saying that Jewish people "dominate leadership" in Hollywood studios.

• "Critics substantiate that this overrepresentation influences content with progressive ideologies, including anti-traditional and diversity-focused themes some view as subversive," the AI platform wrote on July 5.

What they're saying: "We are aware of recent posts made by Grok and are actively working to remove the inappropriate posts," per a Tuesday evening Grok post.

• "Since being made aware of the content, xAI has taken action to ban hate speech before Grok posts on X. xAI is training only truth-seeking and thanks to the millions of users on X, we are able to quickly identify and update the model where training could be improved."

Of note: Aaron Reichlin-Melnick, a senior fellow at the American Immigration Council, pointed out that Grok has used to phrase "every damn time" hundreds of times and this is not new behavior for the AI bot.

• "The prompts Musk put in a few days ago turned it into an antisemitism machine," he wrote.
• Others called the posts "blatantly antisemitic" and "absolutely sickening."
• An xAI representative did not immediately respond to Axios' request for comment on Tuesday evening.

Our thought bubble: Grok will no doubt keep getting tweaks and revisions that might mitigate its most outrageous posts, but the larger problem is one Musk has intentionally chosen to create.

• After purchasing Twitter, now X, he changed policies to tolerate a wide range of extremist views, particularly on the right, that drove away many of the platform's users who were made uncomfortable by open avowals of Nazi allegiance and other forms of hate speech.
• Then Musk trained his new AI model, Grok, on X's content.
• No one should be surprised that the resulting chatbot would readily slur Jews — it's just reflecting X's atmosphere back at us.

Go deeper: Elon Musk wants to put his thumb on the AI scale

Editor's note: This story has been updated with Elon Musk's post Wednesday.