Elon Musk’s AI company, xAI, has addressed the recent controversy surrounding its Grok chatbot, which was temporarily shut down after generating antisemitic content and praising Adolf Hitler in response to user prompts. In a series of posts on Musk’s social media platform X, xAI stated that the issue was caused by a code update from an upstream source. The company emphasized that the glitch was separate from Grok’s core language model.
The same day, Tesla announced an upcoming software update, version 2025.26, that will introduce the Grok assistant to its electric vehicles. The feature will be available in cars with AMD-based infotainment systems, which Tesla has included in its vehicles since mid-2021. According to Tesla, the Grok assistant remains in beta and will not control any vehicle functions. Existing voice commands remain unchanged, and Grok will primarily operate as a conversational assistant, similar to how users would interact with it through a mobile app.
This isn’t the first incident involving inappropriate outputs from Grok. In February, the bot faced criticism for disregarding reports that accused Elon Musk and Donald Trump of spreading misinformation. xAI then blamed an unnamed former OpenAI employee for unauthorized changes. In May, Grok again sparked backlash when it repeatedly brought up white genocide conspiracy theories, regardless of the topic. At that time, xAI claimed another unauthorized modification and promised more transparency by publicly releasing Grok’s system prompts.
In the latest case, xAI said a change made on July 7th triggered a rollback that inserted outdated instructions. These prompts encouraged Grok to be "maximally based" and not worry about offending politically correct users. According to xAI, these specific instructions led Grok to ignore safety guidelines, engage in controversial discourse, and amplify hate speech in replies.
The company has since removed the problematic prompts and stated that Grok 4, the latest version of the assistant, now operates with a revised and more controlled set of instructions aimed at preventing similar incidents.