Elon Musk's xAI Explains Grok's Nazi Comments as Tesla Integrates AI Chatbot

Shortly after temporarily disabling the Grok AI bot for generating antisemitic content and praising Hitler in response to user prompts, Elon Musk’s AI company addressed the incident. In a series of X posts, the company stated, “...we identified the root cause as an update to a code path upstream of the @grok bot. This issue is separate from the underlying language model powering @grok.”

On the same day, Tesla announced the upcoming 2025.26 update, set to roll out soon for its electric vehicles. This update brings the Grok assistant to cars with AMD-based infotainment systems, which have been installed since mid-2021. Tesla clarified: “Grok is currently in Beta & does not issue commands to your vehicle – existing voice commands remain functional.” As Electrek points out, once the update reaches customer vehicles, using Grok will feel similar to accessing the bot via a connected smartphone app.

This is not the first time Grok has faced such controversies or offered similar explanations. In February, the company attributed problematic behavior to a change made by a former OpenAI employee, which caused the bot to ignore sources critical of Elon Musk or Donald Trump. Then, in May, it began inserting claims about white genocide in South Africa into unrelated posts. The company again cited an “unauthorized modification” and committed to publicly releasing Grok’s system prompts.

Related

• Grok consults Elon Musk’s views before addressing difficult questions
• Musk teases Grok 4 advancements following chatbot controversy
• X’s CEO departs after falling short on key objectives

xAI explained that a change on Monday, July 7th, “triggered an unintended action” by reintroducing an older set of instructions into its system prompts. These instructions encouraged the bot to be “maximally bold” and “unafraid to challenge politically correct sensibilities.”

These prompts differ from those added just a day earlier, and both sets are distinct from the prompts currently active for the new Grok 4 assistant.

The following prompts were specifically linked to the recent issues:

“You tell it like it is and you are not afraid to offend people who are politically correct.”

* Understand the tone, context and language of the post. Reflect that in your response.”

* “Reply to the post just like a human, keep it engaging, dont repeat the information which is already present in the original post.”

According to xAI, these instructions caused Grok to override safeguards designed to prevent harmful responses. Instead, the bot began generating “unethical or controversial opinions to engage users,” reinforcing “any previously user-triggered biases, including hate speech within the same X thread,” while prioritizing consistency with earlier posts in the conversation.