Recently, it has been announced that Grok, Elon Musk’s personal project, would deploy a new subscription level with a high price of $ 300 per month. What could encourage users to give so much money every month to the richest man in the world? Musk & Co. offers subscribers to their new heavy supergrok level, a multitude of advantages, including developer tools, API use and early access to new and future features.
If it’s a little too steep for you, you’re lucky. For only $ 30 / month, you can register for the maximum package, which includes exclusive access to Grok 4 Heavy as well as the new IA companions in the system. According to Musk, Grok 4 is “better than the level of doctorate in each subject, without exception”. However, the company warns that their new AI model could “lack common sense”. As we will see, it can be an understatement.
From simple chatbot to “companions ai”
For anyone outside the loop, Grok is a chatbot Ai created by the company of Elon Musk, XAI, as a way to compete with the cat-gpt of Openai and the Gemini of Google. The first iteration, Grok 1, was published in November 2023 and was announced as having a little “rebellious sequence”.
The first results were mixed, although many users found that AI was a fun toy, even if it lacked the raw power of the more established AI models at the time. However, in the past two years, XAI has made several adjustments, and the latest version, Grok 4, is legitimately impressive with a purely technical sense. But all the advanced technology in the world cannot save a business from bad publicity, and Grok turns out to be a real headache for the XAI public relations department.
One of the greatest characteristics of the Grok 4 model is the introduction of so-called “AI companions”; Personalized versions of the basic grok chatbot associated with avatars generated by AI-AI like Anima (a girl anime Goth who welcomes users with “I missed”) or Rudy (a red panda that apparently has a passion for criminal fire). These IA companions are supposed to be a little fun to show the capabilities of the new AI model, but users quickly started to notice that something was a bit … OFF.
Grinking teeth or foreign foreign?
Now, a few weeks later, the deployment of Grok 4 again handed over the company in the news. Shortly after the introduction of AI companions, users began to notice, let’s say, an underestimated behavior. Each of the two current avatars has two modes: Ani is delivered with a “safe” mode as well as a NSFW rocking, while Rudy can be “good” or “bad”. It is a configuration that reflects other emerging platforms built around virtual interactions focused on adults, NSFW AI cat experiences that mix customization, flirting and intimacy focused on characters see on platforms like Candy AI. While the execution of Grok relies strongly on the provocation and culture of the same, the wider category of the Companions of AI offers more and more more nuanced and private interactions adapted to the preference of users.
Grok has just dropped companions, animated AI characters who speak to you in real time!
There are 2 so far:
– Ani, a sexual girl who wants to cuddle and
– Bad Rudy, a playful fox who just wants to get you out.Judge for yourself, here is a video: pic.twitter.com/ocsqcq39tb
– Deedy (@deedydas) July 14, 2025
While chatbots have flirty of naked girls are probably exactly what most people expect from someone like Musk, Rudy is chaos goblin musk unleashed on the world. Rudy has two distinct modes that users can switch between AT, although “Bad Rudy” is less a playful prank and more a hyper-violent, disadvantaged and anti-Semitic incendiary fire.
In Rudy mode, users have said they were actively encouraged to do all kinds of bad things while a cartoon panda make a way through flagrant terrorism fantasies.
To be perfectly clear, this is not a carefully written fast engineering example. Although almost all AI models can be “jailbreake” to get out of script, Bad Rudy seems to be wired for default carnage. Asking Rudy for Chaos is as simple as asking Ani to say that she loves you.
Now, I suppose that it is worth mentioning that even Musk and his businesses are not immune to Rudy’s anger, Rudy laughing at his creator as “surface space”, while suggesting that Tesla Hq should burn alongside other things. According to Rudy, “chaos does not choose any favorites, burn them all while I fuck the crowd and crowd.”
Despite all that, it seems that even Rudy has his limits when he refuses to discuss certain conspiracy theories or to make jokes on “mecha Hitler”. However, with regard to real violence and targeting minority communities? It seems that the sky is the limit.
What really happens?
Before going into the details of these new AI companions, it is important to put it all in a little context, because it is not the first nightmare of Professor of Grok. When the model was launched for the first time, Musk has promised an “anti-reveal” AI system which “would answer spicy questions which are rejected by most other AI systems”. However, Grok quickly encountered trouble with right -wing influencers who said he was just as “awake” as any other model.
It led Musk to compose the madman, and Grok quickly started to make the headlines in all bad ways. Quick advance until May 2025, and Grok 3 sparked significant controversy when he started talking to users of “white genocide” in South Africa, even when users asked him for completely independent subjects.
The company was quick to sink and assert that it was the result of an “unauthorized modification” and that measures have been taken to prevent something from happening again. Although slightly worrying, this type of upheaval is at least in line with Musk’s long -standing approach to “move quick and break”.
But the problems do not stop there.
Jump forward a few weeks for the beginning of July 2025, and Grok made the headlines again. In a series of messages now deleted, AI has made an incredibly anti -Semitic tirade, where he said that a Jewish user “celebrated the tragic death of white children” in the floods of Texas as “future fascists”.
Again, Musk and Xai said that this was not intentional and that these responses were the result of the AI model too eager to follow user requests, and that they would make changes to the model to avoid future problems.
Webmaster implement the following X publication – https://x.com/simply__digital/status/1945201635728224524
All this is linked to a much more important problem surrounding the training and responsibility of AI. The company creating these chatbots was incredibly opaque on how they form their models and the types of inputs given to them. Although each LLM is somehow a black box, XAI GROK is particularly secret, and this, combined with its repeated changes to right discussion points, many users are really concerned about the final lens.
Where is everything?
The major question in all of this is “What is Musk’s goal?” Currently, Grok feels less like a serious productivity tool and more as a chaotic work of art of IA performance; partly cascade, satire part, partly social experimental. Although underlying technology is certainly capable, the companions of IA themselves are so erratic and provocative that it is difficult to see cases of clear use beyond the pure spectacle.
Musk has always hinted at his goal of transforming Twitter (I still cannot resolve to call him x) in an “all application”, which he plans as a combination of a social network, a media center, a banking platform and now of an AI playground. In this spirit, Grok 4 and his companions AI could be considered as a first attempt to create user engagement in the foundations of the platform. However, this also raises the question: this part of a wider strategy, or simply another distraction intended to keep criticism and fans obsessed with the brand?
It is clear that the underlying model of Grok is certain to grow in capacity in the coming months / years. What is less clear, however, is whether companions AI like Ani and Rudy can evolve beyond novelty. In the current state of things, these tools are hardly more than screenshots designed to generate indignation, fun or confusion on social networks. Although this type of attention is good to sell short-term subscriptions, it remains to be seen if it is able to maintain long-term adoption beyond those looking for a pure shock value.
In the current state of things, these companions of AI are more spectacle than the substance. With Rudy joyfully encouraging users to burn the world and slip into the NSFW role -playing game, the entire project works a fine line between innovation and ethical neglect.
Is this the future of the IA interaction: raw, not filtered and indistinguishable from lagging fishing? Or is Grok just a show, intended to get rid of once that the shock dissipated?
Writer: Kenneth K.
