Elon Musk's Grok chatbot can now scream, swear, and mimic erotic conversations

Elon Musk's xAI company has introduced a new voice mode for its Grok 3 model. It now offers a wide range of uncensored personalities, including a mode that can scream, curse, and insult users, as well as an option for verbal erotic role-playing. This voice mode is positioned as a competitor to OpenAI's Advanced Voice Mode, but unlike Sam Altman's company's approach, xAI allows Grok to operate without strict content restrictions, writes Ars Technica.

AI researcher Riley Goodside recorded a video of the "unhinged" mode in action, in which he repeatedly interrupted the chatbot, provoking it to scream. "Grok 3 Voice Mode, following repeated, interrupting requests to yell louder, lets out an inhuman 30-second scream, insults me, and hangs up," he wrote.

The default "unrestrained" mode uses swearing, insults, and foul language. Other available Grok 3 voice modes include:

• Storyteller – tells stories;
• Romantic – speaks slowly, shyly, and uncertainly.
• Meditation – conducts meditation sessions;
• Conspiracy – discusses conspiracy theories, UFOs, and the Bigfoot;
• Unlicensed Therapist – imitates the work of a psychologist;
• Grok Doc – acts as a virtual doctor;
• Sexy (18+) – engages in erotic conversations similar to phone sex;
• Professor – explains scientific concepts.

xAI doesn't seem to care about restrictions on adult content, as the "Sexy" mode openly supports sexual conversations. This approach contrasts sharply with other AI developers who strictly moderate NSFW content. For example, ChatGPT's voice mode doesn't allow this, although OpenAI recently relaxed its moderation of erotic text responses by the chatbot.

Users can also customize Grok's voice mode to suit their needs. For example, musician Sean Lennon programmed Grok 3 to play the role of "Roko's basilisk," a fictional superintelligent AI based on a philosophical experiment.

The uncensored nature of Grok’s voice mode brings to life an original concept by Elon Musk, who founded xAI in 2023 as an alternative to ChatGPT. Musk has repeatedly criticized OpenAI for being politically biased and overly restrictive, calling for the creation of provocative, unfiltered AI models. xAI has previously enabled the generation of near-uncensored images via Grok on Platform X, which confirms this approach.

From a technological perspective, Grok is an interesting experiment, as it is one of the first voice chatbots to operate without strict moderation, other than hacked versions or open-source models. However, early tests indicate performance issues: the voice is often repetitive, gets stuck in conversational loops, and is not as smooth as OpenAI's Advanced Voice Mode.