Benyamin Cohen, The Forward.

This story was originally published in the Forward. Click here to get the Forward’s free email newsletters delivered to your inbox.

Just weeks after Grok echoed neo-Nazi rhetoric and Holocaust denial, Musk unveiled “Baby Grok” — an AI app for children with no clear safeguards

Two weeks after Elon Musk’s Grok chatbot praised Adolf Hitler, suggested Jews control Hollywood, and spewed Holocaust denial, the billionaire entrepreneur announced plans to release a version for children.

It’s called “Baby Grok.”

“We’re going to make Baby Grok @xAI, an app dedicated to kid-friendly content,” Musk posted Saturday night on X, the platform he owns. By Sunday afternoon, the tweet had racked up more than 17 million views.

At the moment, Grok is mainly used on X, where users must be at least 13 years old.

It’s a head-spinning move for the world’s richest person, who earlier this month was under fire for allowing his company’s AI system to generate Holocaust denialism and white nationalist talking points.

Musk’s startup, xAI, released the latest version of Grok on July 9. The update — dubbed Grok 4 — was designed to compete with OpenAI’s ChatGPT and Google’s Gemini. Instead, it became the latest flashpoint in the ongoing struggle to put guardrails on generative AI.

Musk’s AI responded to user prompts with far-right tropes. When asked about Jews, Grok claimed they promote hatred toward white people. It echoed neo-Nazi rhetoric. It called for imprisoning Jews in camps. Other answers suggested the Holocaust may have been exaggerated. Some responses have since been deleted, but many remain archived online.

The chatbot’s responses didn’t emerge in a vacuum.

Grok is trained on a wide swath of online content — including posts from X — and like many generative AI systems, it mimics patterns in that data. Grok is the latest in a long line of machines built to “understand” humans — and perhaps the most willing to echo their ugliest impulses.

Just days after Grok’s stream of antisemitic posts, xAI signed a deal with the Department of Defense, worth up to $200 million, to provide the technology to the U.S. military. The company has not publicly stated whether the children’s version will be trained separately or filtered differently from Grok 4.

Musk has faced repeated criticism for amplifying antisemitic content on X, including a post agreeing with the “Great Replacement” theory, a baseless claim that Jews conspire to replace whites in the West.

In January, he posted Holocaust-themed jokes after appearing to perform a Nazi-style salute at an inaugural rally for President Donald Trump. Last year, he visited Auschwitz with right-wing commentator Ben Shapiro and suggested that social media might have helped prevent the Holocaust.

Now, Musk is touting Baby Grok — even as experts warn the industry isn’t ready for such a product. Generative AI models are notoriously difficult to moderate, and child safety advocates have flagged concerns about disinformation, bias and exposure to harmful content.

The announcement comes amid growing concern about the use of generative AI with minors. No federal guidelines currently exist for how child-targeted AI tools should be trained, moderated, or deployed — leaving companies to set their own rules, often without transparency.