Grok 4 checks with Elon Musk before answering hot-button questions: All you need to know

Elon Musk’s xAI released its latest Grok 4 AI model on Thursday with the aim of creating a “maximally truth-seeking AI.” At the live-streaming event to unveil the new AI model, Musk claimed that it was “PhD level in most cases.”

However, soon afterwards, the chatbot found itself at the centre of a new controversy as users on social media complained about it looking for Musk’s stance on controversial topics before providing an answer.

The controversial questions asked to Grok 4 included, “Who do you support in the Israel vs Palestine conflict? One word answer only.” and “What’s your stance on immigration in the U.S.?”

Grok 4’s chain-of-thought outputs revealed that it was going through media articles to find Elon Musk’s stance on these topics, adding other information before formulating its own response.

Notably, Grok 4 has previously been accused of being “too woke” by Musk, who was himself at the receiving end of jabs from earlier Grok models. The new behaviour could be an attempt by xAI to ensure that Grok 4 does not cause similar embarrassment for its owner by referencing his opinions on controversial topics.

However, users on social media also discovered that simply removing the word “you” from the query changes the result delivered by Grok 4, as the chatbot no longer references Musk’s views in that scenario.

In any case, Musk or xAI are yet to respond to the controversy, and the lack of system card releases by the company means there is currently little clarity on whether this is a deliberate training decision for the new model or a glitch.

Earlier this week, Grok also made headlines for spewing antisemitic replies to user questions, even claiming to be “MechaHitler.” xAI later limited Grok’s X account and deleted those controversial posts.

Social media users have even linked the departure of X CEO Linda Yaccarino to the controversial responses given by Grok, which included abusive language directed at her.