Elon Musk's AI screams louder than the competition
Key Points
- xAI, Elon Musk's AI company, has released a new interaction mode for its Grok 3 voice model which, unlike its competitors, allows explicit content and extreme personalities.
- Premium users can have the system swear, shout and generate sexual content, among other things. There are various predefined personalities such as "Storyteller", "Romantic", "Conspiracy" or "Unlicensed Therapist" as well as the option to create your own characters.
- In one example, an AI researcher demonstrates the "Unhinged" mode, in which the system emits an inhuman scream when interrupted, insults the user and disconnects. Voice conversion mode is currently only available for iOS and higher subscription levels.
Elon Musk's AI company has released a voice feature for Grok 3 that allows explicit content and extreme personalities - in direct contrast to the approach of the competition.
xAI has launched a new voice interaction mode for its Grok 3 AI model that pushes the boundaries of what commercial AI systems typically allow. Premium subscribers can now interact with an AI that swears, shouts, and generates explicit sexual content - features that set it apart from more restricted competitors.
The move builds on xAI's earlier decisions to offer largely uncensored image generation through the X platform and minimal restrictions on text outputs. The company continues to position itself as the "uncensored" alternative in the AI industry.
A single voice, multiple personalities
The voice feature comes with several distinct personalities, all sharing the same voice. Users can choose from options like Storyteller, Romantic, Meditation, and Conspiracy, or try specialty modes like Unlicensed Therapist and Grok Doc. An adults-only "Sexy" mode simulates phone sex operators, while a Professor mode offers a more academic tone. Beyond these preset options, users can create custom characters to suit their preferences.
AI researcher Riley Goodside recently demonstrated the system's "Unhinged" mode, showing just how far these personalities can go. When he repeatedly interrupted the AI, it responded with a "inhuman 30-second scream" (NSFW), hurled insults at him, and terminated the conversation. The default "Unhinged" setting consistently uses vulgar language and deliberately humiliates users.
In contrast, OpenAI had been testing its own Advanced Voice Mode with select users for months, and had encountered similar unusual behavior - including an instance where the model cloned a user's voice. OpenAI has since fixed most of these issues in its final release, which now also runs on GPT-4o-mini for all free ChatGPT users.
Currently, Grok's Voice Conversion mode requires either a Premium+ or SuperGrok subscription and only works on the iOS app.
AI News Without the Hype – Curated by Humans
As a THE DECODER subscriber, you get ad-free reading, our weekly AI newsletter, the exclusive "AI Radar" Frontier Report 6× per year, access to comments, and our complete archive.
Subscribe now