Don’t use Grok’s new voice feature at work, definitely not in “sexy” mode

Grok 3 now has a voice, and its not safe for work (NSFW) chats have the internet in a chokehold.

Elon Musk posted on X that xAI’s model Grok 3 will offer a voice mode Premium+ and SuperGrok users.

Try Grok voice conversation mode!

Requires a Premium+ or SuperGrok subscription.

pic.twitter.com/247Ev60DoJ

undefined Elon Musk (@elonmusk) February 24, 2025

It appears that Grok is trying to mimic OpenAI’s “Advanced Voice Mode,” but with a twist.

Instead of declining NSFW requests or unhinged chats like ChatGPT does, the Grok 3 model actually welcomes the chaos.

In a video found by Ars Technica, a video shows Grok’s NSFW mode, which shows the AI model cursing, calling the user a “bitch” and screaming other expletives.

When prompted, the AI model just started screaming, crying out in an inhumane and animalistic way.

The X user said, “Grok 3 Voice Mode, following repeated, interrupting requests to yell louder, lets out an inhuman 30-second scream, insults me, and hangs up.”

Grok 3 Voice Mode, following repeated, interrupting requests to yell louder, lets out an inhuman 30-second scream, insults me, and hangs up pic.twitter.com/5GtdDtpKce

undefined Riley Goodside (@goodside) February 24, 2025

Grok’s voice mode has various settings, including storyteller, romantic, unhinged, Grok “Doc,” unlicensed therapist, and “sexy,” which is designed for 18+ users.

Supposedly, the “unhinged” model insults, belittles, and curses, using profanities and lewd language to interact with the user, Ars Technica reports.

“Romantic” mode interacts with users in a “slow, uncertain, and insecure way,” while “meditation” mode guides you through a meditative experience as a tonic to the chaos.

Don't miss our latest stories on Google News

Follow us

There’s even “conspiracy” mode, where the voice tells you about various conspiracy theories ranging from bigfoot to unidentified aerial phenomena (UAPs).

“Sexy” mode has been described as a modern-day phone sex hotline where your AI operator can whisper dirty nothings in your ear.

While these modes are typical of Musk’s AI model, as previously Grok was known as the most unhinged chatbot around, they’re not commonplace in the chatbot market, where many AI models are censored and NSFW content is not tolerated.

Despite being the supposedly uncensored answer to ChatGPT, Grok temporarily refused to provide users with sources that mentioned how Elon Musk and Donald Trump spread misinformation,” The Verge reported.