Grok’s New AI Voice Mode Lets It Swear, Scream, Insult, and Have Explicit Conversations

Elon Musk’s artificial intelligence company, xAI, has launched a voice mode for its Grok chatbot that allows it to swear, yell, and simulate intimate conversations.

Unlike AI assistants such as ChatGPT, Google’s Gemini, and Microsoft Copilot, which adhere to strict content moderation, Grok’s design prioritizes free speech over conventional AI safety measures.

The launch of this feature reinforces Musk’s long-standing criticism of AI censorship and his push for a chatbot that delivers responses with fewer restrictions. However, it has also raises ethical concerns, particularly around AI-generated misinformation, emotional manipulation, and the lack of defined content moderation policies.

How Grok’s Voice Mode Breaks AI Norms

AI voice assistants have become a competitive focus in the industry, with OpenAI expanding ChatGPT’s capabilities to include live video input and Google introducing Gemini Live for real-time AI conversations. Microsoft has just removed AI usage limits on Copilot for free users including unlimited AI-driven voice interactions to allow broader public access.

Unlike these competitors, Grok takes an unconventional approach by , offering an “unhinged” mode, which is set to aggressively insult, curse, and mock the user continuously with crude language.

Other available modes include “Storyteller,” which focuses on narrating tales, and “Romantic,” where the AI speaks hesitantly, with a slow and uncertain tone.

“Meditation” offers a guided relaxation-like experience, while “Conspiracy” engages in discussions about UFOs, Bigfoot, and various conspiracy theories. “Unlicensed Therapist” takes on the role of a conversational psychologist, and “Grok Doc” mimics a medical professional.

There’s also “Sexy,” labeled as “18+,” which resembles an adult hotline operator, and “Professor,” which delivers explanations on scientific topics.

AI researcher Riley Goodside highlighted the extreme nature of “unhinged” mode in a video shared on X. In the clip, he repeatedly interrupted the chatbot, prompting it to escalate into simulated shouting when asked.

Other shared videos show how Grok in “unhinged” mode curses, insults, and belittles non-stop with vulgar language during conversations. This level of expressiveness, which includes profanity and roleplay-like conversations, is not permitted in OpenAI or Google’s AI models.

While OpenAI’s ChatGPT Voice Mode and later Advanced Voice Mode was designed to improve conversational flow within ethical guidelines, Grok removes many of these restrictions. This makes it the most expressive AI chatbot available, but it also raises concerns about where AI speech boundaries should be drawn.

Ethical Concerns and the Debate Over AI Moderation

The removal of AI speech restrictions has reignited debates over whether chatbots should prioritize unrestricted expression over responsible content moderation.

AI developers have historically implemented safeguards to prevent chatbots from generating harmful, misleading, or inappropriate content. OpenAI and Google, for example, have enforced strict behavioral limitations in their AI models, even at the risk of making interactions feel overly sanitized.

Musk, however, has repeatedly criticized these restrictions, framing them as ideological constraints rather than ethical safeguards.

While Grok’s approach may appeal to those who prefer a more unfiltered AI, it also raises questions about the long-term implications of removing content moderation.

Critics worry that this could enable problematic AI interactions. Without clear guidelines on where Grok draws the line, concerns persist about the chatbot’s potential role in spreading misinformation, engaging in manipulative conversations, or responding in ways that could be deemed offensive or harmful.

OpenAI has explicitly stated that its AI models will not engage in certain discussions or use language deemed inappropriate, while Google has taken similar steps to ensure Gemini remains within predefined ethical boundaries.

Grok 3 Moves Behind a Paywall

xAI has decided to keep Grok 3 locked behind the X Premium+ subscription, which recently doubled in price to $40 per month. The decision aligns with Musk’s broader push to shift X (formerly Twitter) toward a subscription-driven model.

By contrast, OpenAI offers its AI services under tiered pricing models, with ChatGPT Plus ($20/month) and ChatGPT Pro ($200/month). Meanwhile, Microsoft’s strategy involves integrating Copilot into its software products, making its AI features more widely accessible.

By tying AI access to its highest-tier subscription, xAI is positioning Grok as an exclusive product rather than a widely available tool. Whether this strategy succeeds depends on whether users are willing to pay for an AI that prioritizes free speech over content moderation.

Does Grok 3 Compete Technically With OpenAI and Google?

Beyond its voice capabilities, xAI claims that Grok 3 is technically competitive in key AI benchmarks, particularly in math, coding, and logical reasoning.

According to internal tests Grok 3 scored 52 in Math (AIME’24), significantly ahead of GPT-4o (9) and Claude 3.5 Sonnet (16). In Science (GPQA), it led with 75, outperforming Gemini 2 Pro, Claude 3.5, and DeepSeek-V3, which all scored 65, while GPT-4o lagged at 50.

While these benchmarks suggest Grok 3 excels in structured reasoning, independent testing has yet to validate xAI’s claims. The model runs on the Colossus supercomputer, which Musk is scaling to one million Nvidia GPUs to compete with OpenAI and Google’s AI infrastructure.

A notable feature of Grok 3 is its Think button, which allows users to request more complex and structured responses.

According to xAI, “The Think button enables advanced chain-of-thought reasoning, which like OpenAI’s o1 and o3 models and DeepSeek R1 aims to provide users with results based on complex thinking.”

This mirrors similar functionalities in Microsoft’s Think Deeper feature in Copilot, Perplexity’s implementation of reasoning models such as OpenAI’s o3-mini and DeepSeek R1, and in Anthropic’s new Claude 3.7 Sonnet reasoning model which distinguishes itself by allowing to set constraints on response time.

Where AI Voice Assistants Are Headed Next

Grok 3’s voice mode is not just an AI feature—it signals a broader shift in how AI models are being developed. While OpenAI and Google continue refining their systems for accuracy, ethical safeguards, and real-time responsiveness, xAI is actively marketing the idea that AI should not be restricted by traditional moderation policies.

Meanwhile, AI capabilities are rapidly expanding beyond text and voice. OpenAI recently introduced live video analysis in ChatGPT’s Advanced Voice Mode, allowing users to interact visually with the AI while talking to it.

Google has been improving Gemini Live’s ability to analyze user-provided files in real time, increasing its versatility. 

Unlike these efforts, which are focused on expanding AI’s practical applications, xAI’s bet on unrestricted AI speech represents a “philosophical” departure. Instead of emphasizing use cases in education, enterprise, or research, Grok’s voice mode is largely being marketed as a personality-driven alternative that offers a different conversational experience.

Musk’s AI Strategy: A Risky Experiment?

By tying Grok’s most advanced features to the highest-tier X Premium+ subscription, Musk is taking a different approach from competitors that are expanding access to AI rather than restricting it. 

This raises an important question: Is Grok’s unfiltered voice mode a genuine competitive advantage, or simply a way to generate interest in X’s paid subscription model?

While Musk has positioned xAI as a disruptor in the AI space, Grok still remains significantly behind ChatGPT and Gemini in real-world adoption and enterprise integrations.

For now, Grok remains an outlier in the AI assistant market. Its ability to yell, swear, and roleplay will appeal to a niche audience, but it is unlikely to gain widespread adoption in professional settings, where companies prioritize AI safety and ethical constraints over expressive AI behavior.

If unfiltered AI speech proves to be more of a novelty than a necessity, Grok’s appeal may remain limited compared to more structured AI models.

Table: AI Model Benchmarks – LLM Leaderboard 

Last Updated on March 3, 2025 11:27 am CET