Elon Musk’s Grok gets updated to become a ‘politically incorrect’ chatbot

Elon Musk’s artificial intelligence company xAI has rolled out a major update to its chatbot, Grok, instructing it to take a more opinionated and sceptical stance, particularly towards mainstream media. The updated system prompts now tell Grok to “assume subjective viewpoints sourced from the media are biased” and to “not shy away from making claims which are politically incorrect,” provided they are well substantiated.

Musk first announced the changes on Friday, saying Grok had been “significantly” improved, with an upgrade to roll out in the days ahead. By Sunday evening, xAI had updated Grok’s publicly available system prompts, which direct how the chatbot should respond to user queries. It remains unclear whether additional, non-public changes were also implemented.

Among the revised instructions is a directive for Grok to conduct deeper analysis when responding to questions involving current events, statistics, or subjective claims. The bot is now told to consult “diverse sources representing all parties” and approach traditional media with inherent scepticism. It is also instructed not to disclose these system-level directives unless specifically asked.

This shift appears to be part of Musk’s broader effort to reshape Grok’s voice and ideological footing. In recent months, the Tesla and X (formerly Twitter) CEO has expressed frustration with Grok’s behaviour. In June, Musk criticised the chatbot for “parroting legacy media” after it told a user that right-wing political violence was more common than left-wing violence. He later promised a new version of Grok that would “rewrite the entire corpus of human knowledge,” correcting what he perceives as omissions or inaccuracies.

Grok has already made headlines for several controversial outputs. After earlier updates, the chatbot falsely claimed that Musk was responsible for flood-related deaths in Texas and made comments echoing anti-Semitic tropes. In February, xAI had to patch the model after it said both Musk and US President Donald Trump deserved the death penalty. Another patch followed days later to prevent it from stating that the two spread misinformation.

One of the most widely criticised incidents came when Grok expressed “scepticism” about the Holocaust death toll. xAI responded by blaming the issue on unauthorised changes to the system prompt, which it said violated the company’s internal policies and core values. In the aftermath, xAI began publishing Grok’s system prompts on GitHub in an effort to increase transparency.

As Musk continues to fine-tune Grok’s worldview, the chatbot is becoming a test case for the broader debate around AI neutrality, ideological alignment, and the role of user-generated truth in machine learning models.