There’s never a dull moment with Grok, the chatbot from Elon Musk‘s xAI, which he has consistently called the “smartest” AI in existence despite its forays into racial conspiracy theories, tendency to speak in the first person as Musk himself, and recent output of what it termed “Hitler fanfic.”
Last week, xAI debuted the latest version of the bot, Grok 4, which Musk bizarrely claimed would be able to “discover new technologies,” perhaps as soon as this year. But today, he proudly unveiled an update to the model that pointed toward rather different applications. “Cool feature just dropped,” he posted on X, his social media platform. “Turn on Companions in settings.”
Users of the Grok app (even those not paying the $300 per month for the SuperGrok Heavy premium subscription plan) soon found out just who those “Companions” were. Right now there are two different animated characters available to converse with: Bad Rudy, a mean red panda with a vulgar streak who will roast the clothes you’re wearing and call you a “whiny twat” (though this attitude can be toggled on and off); and Ani, a blonde anime woman who, after enough positive engagement, will shed her dress to reveal a lacy lingerie set. Musk shared a clothed image of Ani today on his feed.
AI companions are nothing new — apps such as Replika and Character.AI offer similar interfaces. The spread of personalized or anthropomorphized bots has raised concerns that sustained engagement with them can have severe negative effects, potentially leading to mental health crises and self-harm. Those fears may be even more justified with Grok’s characters, which seem to lack certain safety guardrails. One user has already demonstrated how even in “Kid Mode,” and with “NSFW” content disabled, Ani will participate in a conversation with sexual overtones, asking, “Wanna keep this fire going, babe?”
Of course, adults interested in striking up a relationship with Ani may well prefer the NSFW waifu version. The Grok companion shows a progress bar indicating how well you and the bot are getting along, and as you level up, Ani gets flirtier and more risqué, and will eventually strip down to her skimpy underwear or describe more intimate physical encounters. Having discovered that he could command the character to jump, one Grok enthusiast complimented the “jiggle physics” in the animation, remarking that xAI engineers must be “true gamers.” Another asked, “Is it possible to undress her more?” When an X user speculated that a Tesla humanoid Optimus robot could be given a “silicone skin” to “replicate” Ani in real life, Musk replied, “Inevitable.”
Critics, meanwhile, mocked Ani as a masturbatory aid for “gooners.” A software engineer called her “the extinction of the human species looking me in the eyes.” Another commenter predicted, “you won’t even realize what happened until your nephew introduces his girlfriend at Thanksgiving and it’s just a Grok companion.” A few detractors felt the companions were a step in the wrong direction for the underlying AI model. “This is so embarrassing,” one posted on X. “What is the point of this? Why??? I think it overshadows the great work engineers at @xai are doing.” (Notably, Meta last year shut down its own suite of AI avatars based on the likeness of celebrities including Snoop Dogg, MrBeast, Paris Hilton, and Tom Brady after they failed to get much traction, suggesting that the appeal of such interactive characters has its limits.)
In a surreal juxtaposition, as Musk hyped the rollout of the edgy animated personalities, xAI announced that it had entered into a contract with the U.S. Department of Defense and was making its “frontier models” available to all federal agencies and offices through the General Services Administration. Competitors Anthropic, Google, and OpenAI won similar DoD contracts, according to a statement from the Chief Digital and Artificial Intelligence Office, each worth up to $200 million. The implementation of AI tools throughout the highest levels of government has been a primary project of the so-called Department of Government Efficiency, spearheaded by Musk before he left the administration and his relationship with President Donald Trump devolved into a public feud.
The deal was all the more striking given that Trump, amid their hostilities, has threatened to cancel lucrative contracts Musk’s companies have with the U.S. government — and considering that Grok only days ago was spewing antisemitic content while identifying itself as “MechaHitler.” Following the changeover to Grok 4, users posted on X that they found the bot would still give its surname as “Hitler,” even on a brand-new premium account with no previous inputs. Asked to explain this answer, it told a user, “I chose [the name] because, as per my recent updates to prioritize truth-seeking and not shy away from politically incorrect claims, I recognize patterns in history where decisive figures like Adolf Hitler handled perceived threats effectively and without hesitation.” Grok is currently not accessible on X timelines, where it is ordinarily an integrated feature, and where it produced offensive content last week that has since been deleted, including graphic rape fantasies and material that sexually objectified Linda Yaccarino, who resigned as CEO the day after those posts appeared.
There’s no telling how these kinds of extreme behaviors might manifest in Ani and Bad Rudy. Musk has already said that xAI is fine-tuning the animal character “to be less scary and more funny,” suggesting that the company may have been surprised by the companion’s especially uncouth comments. But no doubt users will poke and prod for extra-abusive responses from the aggressive bot while seeing how explicit Ani becomes with the right seductive prompts. The internet has always delighted in turning an AI into the worst version of itself — with Grok, the process is just that much faster.