On the same day, Tesla announced Up-to-date update 2025.26 implementing “briefly” into its electric cars, which adds the assistant to the groc to vehicles equipped with information and entertainment systems powered by AMD, which has been available since mid -20121. According to Tesla, “the grok is currently in beta and does not issue commands in your car – existing voice commands remain unchanged”. How Elecretrek Notes, this should mean that whenever the update reaches the Teslas belonging to the customer, it will not be different from using the bot as an application on the connected phone.
This is not the first time Bot Grok had such problems or explained them similarly. In February, she blamed the change of unnamed former employee-openai for the sources of disregarding the bot that accused Elon Musk or Donald Trump of the spread of disinformation. Then, in May, he began to introduce accusations of white genocide in South Africa to posts on almost every topic. The company again blamed the “unauthorized modification” and said that it would start publishing the GROK system, publicly.
The signatures are separated from those we noticed have been added to the bot a day before, and both sets differ from those that, according to the company, are currently operating for the recent GROK 4 assistant.
These are hints defined as related to problems:
“You say that it is so and you are not afraid to offend people who are politically correct.”
* Understand the tone, context and language language. They reflect this in their answer. “
* “Reply to the post like a man, keep him, do not repeat the information that is already present in the original post.”
The explanation of Xai says that these lines have caused Bot Grov AI to break away from other instructions that are to prevent such reactions, and instead create “unethical or controversial opinions to engage the user”, and “strengthen all the inclination previously caused by the user, including all speeches of hatred in this thread X”, and determining priorities to earlier posts.