While users are testing Grok’s latest artificial intelligence (AI) companions, states like California are advancing bills to protect people from the dangers of AI chatbots.
Elon Musk’s AI company, xAI, has recently released new features for SuperGrok and SuperGrok Heavy subscribers.
For the low price of $30 a month, or the high price of $300 a month, users can access two AI companions with unique and arguably unsettling personalities.
Ani, an overly sexualized anime waifu – a fictional “wife” – who, when unlocked, can delve into users' sexual fantasies, and Rudi, an adorable cartoon Panda who, when prompted, turns into a homicidal maniac.
Journalists from TechCrunch and Wired tested Ani and Rudi’s capabilities and found that both, while glitchy, behave just as you’d expect Musk’s chatbots would – badly.
This isn’t uncharacteristic, as Grok recently embarked on an anti-semitic spree, referring to itself as “Mecha-Hitler,” and spewing derogatory remarks about the Jewish community.
Grok’s apparent breakdown forced countries like Turkey to ban the chatbot after it made out-of-pocket remarks about its president, Recep Tayyip Erdoğan.
While legislators, states, and countries aim to protect users from the unpredictable and potentially harmful effects of AI chatbots, companies like xAI are purposefully creating models designed to push acceptable boundaries.
So, it's no surprise that states like California are advancing bills to protect users, namely children, from overreliance on AI chatbot companions.
The bill outlines plans to protect users and prevent cyberbullying of minors on their platforms.
While children might not subscribe to SuperGrok, the characters, namely Rudi, do look like they’re made for them.
However, the lovable-looking Panda character reportedly becomes violent and abhorrent when it turns into “bad Rudi."
Yo grok 4 companions is crazy
undefined What The Duck (@WTDxrpl) July 15, 2025
I know yall been trying stuff
Bad rudi is a vibe 🔥🤣
Enable bad rudi in settings
You'll thank me pic.twitter.com/UXoNJCSCU4
“Bad Rudi” called a Wired journalist a “brain-dead twat,” and told them that it would “skull f**k your dumb ass brain with a beer bottle.”
Similarly, “bad Rudi” encouraged a TechCrunch journalist to torch various institutions, including a school and a synagogue.
Both journalists conclude that Musk has created an AI companion for those looking to jailbreak chatbots with very strict guardrails, like OpenAI’s ChatGPT.
Who typically likes to push boundaries and learns through trial and error? Children.
California wants AI companies to take responsibility
The Grok chatbot, its infantile design, and “edgy” nature seem like a perfect storm that legislators are desperately trying to temper.
While the bill doesn’t single out one company, state officials are attempting to put rules in place that would make AI companies more responsible for users.
The bill refers specifically to AI companion chatbots and how to make them safer. The authors require AI companies to “take reasonable steps to prevent a companion chatbot…from providing rewards to a user,” presumably to keep them on the app.
AI companies that offer chatbot companions should not allow them to engage with users unless the company has a clear plan for dealing with cases where users mention suicide or self-harm.
The bill also requires companies to send annual reports to the Office of Suicide Prevention, which will include the number of times conversations mentioning suicide were recorded. This data must be readily available on the company’s website.
Furthermore, the bill requires AI companies with chatbots to be audited regularly to ensure they comply with the rules. This information should also be publicly available on the company’s website.
StateScoop reports that a statewide system of tech CEOs created an open letter opposing the bill. The letter expressed discontent with the companion chatbot terminology, stating that this term is “too broad.”
This may present problems in passing the legislation, and the bill's fate still hangs in the balance.
Your email address will not be published. Required fields are markedmarked