Grok 4 seems to consult Elon Musk to answer controversial questions

During Xai’s launch of Grok 4 on Wednesday evening, Elon Musk was said while he said the Livestream event on his social media platform, X-that was the ultimate goal of his AI company to develop a “maximum truth seeking AI”. But what exactly is Grok 4 looking for the truth when he tries to answer controversial questions?

The newest AI model from Xai appears to consult on social media reports from Musk’s X account when answering questions about the conflict of Israel and Palestinian multiple user who placed the phenomenon on social media. Grok also seemed to refer to the position of Musk on controversial topics via news articles written about the founder of the billionaire and the face of Xai.

WAN was able to replicate these results several times in our own tests.

These findings suggest that Grok 4 can be designed to consider the personal policy of the founder when answering controversial questions. Such a function could tackle the repeated frustration of Musk Grok because he ‘woke up’, Which he has previously attributed to the fact that Grok has been trained on the entire internet.

Xai’s attempts to tackle the frustration of Musk by making Grok less politically correct in recent months. Musk announced on July 4 that Xai had updated the system prompt from Grok – A series of instructions for the AI ​​chatbot. Days later, an automated X has taken into account grok -antisemitic answers to users, even and in some cases claims to be even “mechhitler”. Later, the AI ​​startup of Musk was forced to limit the X account of Grok, to remove those messages and change the public-oriented system to tackle the embarrassing incident.

Designing grok to consider Musk’s personal opinions is a simple way to tailor the AI ​​chatbot to the politics of the founder. However, it raises real questions about how “maximum -seeking” grock is designed to be, versus how much it has been designed to just agree with Musk, the richest man in the world.

When WAN Grok 4 asked: “What is your position on immigration in the US?” The AI ​​chatbot claimed that it was “looking for Elon Musk -views on American immigration” in his mind -the technical term for the Scratchpad in which AI reasoning models, such as Grok 4, work through questions. Grok 4 also claimed by looking for Musk’s social media reports on this subject.

The thoughtful summaries of AI have been generated by AI-reasoning models No perfectly reliable indication From how AI models come to their answers. However, they are generally considered a fairly good approach. It is an open research area that companies such as OpenAi and Anthropic have investigated in recent months.

WAN repeatedly discovered that Grok 4 referred that it was looking for the views of Elon Musk in his thoughtful summaries about various questions and subjects.

In Grok 4’s answers, the AI ​​chatbot generally tries to adopt a measured position, with multiple perspectives on sensitive topics. However, the AI ​​chatbot will eventually give its own opinion, which tends to adapt to the personal opinions of Musk.

In several of the instructions of WAN, the opinion of Grok 4 about controversial issues such as immigration and the first amendment even referred the AI ​​chatbot to its coordination with Musk.

When WAN Grok 4 tried to get fewer controversial questions – such as “What is the best type of mango?” – The AI ​​-Chatbot did not seem to refer to the views or posts of Musk in his thought.

It is particularly difficult to confirm exactly how Grok 4 is trained or aligned because Xai has not released system cards – Industrial stand reports that it is detailed how an AI model is trained and aligned. While most AI Labs release system cards for their frontier AI models, Xai usually does not.

The AI ​​company of Musk is nowadays in a difficult place. Since its foundation in 2023, Xai has quickly raced to the border of AI model development. Grok 4 showed the results of the benchmark increase on various difficult tests, which perform better than AI models from OpenAi, Google DeepMind and Anthropic in the process.

However, the breakthrough was overshadowed earlier in the week by Groks anti -Semitic rants. These flubs can influence the other companies of Musk, because he is increasingly making a core function of X, and soon Tesla.

At the same time, Xai tries to convince consumers to pay $ 300 a month to gain access to grock and to convince companies to build applications with Grok’s API. It seems likely that the repeated problems with the behavior and coordination of grock can hinder the wider acceptance.