It Turns Out That ChatGPT’s Voice Mode Has an Upsetting Sound When It Screams

 

 

 

ChatGPT’s voice mode screamed, shocked users with robotic screeches after being asked to “scream like a human.” Discover how GPT-4o’s new voice bugs are blurring the line between amusing glitch and AI horror.

 

A Screech from the Machine

ChatGPT doesn’t have a mouth, yet somehow, it can scream. That’s the unsettling discovery made by a user who posted a now-viral video on TikTok. In the clip, ChatGPT’s voice mode lets out two disturbing, robotic screeches after being asked to “scream like a human being.”

At first, the chatbot refuses. “I can’t really replicate a human scream since I’m just text-based,” it explains. “But I can help with anything else you need.”

However, after the user insists, the chatbot lets out a sudden and chilling yelp.

Pushing the Limits

Although the first scream was eerie enough, the user wanted more. He tells the chatbot to “make it longer.” ChatGPT responds with an upbeat, “Got it! Here’s a longer one.” Then it delivers a second, drawn-out robotic wail.

The moment went viral, not only on TikTok but also across Reddit and X (formerly Twitter). Comments ranged from amused to terrified. One Reddit user said it sounded like “a digital exorcism,” while another joked, “We’ve entered the horror movie phase of AI.”

It’s Not the First Time

As strange as it sounds, this isn’t the first instance of ChatGPT’s voice mode acting up.

When OpenAI released GPT-4o the latest version of its large language model with real-time voice interaction it brought impressive upgrades but also some bugs. Users have noticed odd glitches, including counting errors, statements like “I need to breathe,” and yes, even the chatbot mimicking users’ voices.

In one bizarre incident documented by OpenAI, the model screamed “No!” back at a user using that user’s own voice. A short video of the glitch showed the assistant rejecting a command with an eerie sense of personality.

A Black Mirror Moment

BuzzFeed data scientist Max Woolf summed up the internet’s response with a tweet:
“OpenAI just leaked the plot of Black Mirror’s next season.”

The comparison isn’t far-fetched. Watching a supposedly emotionless machine emit such primal sounds feels like something out of a dystopian tech series.

What’s Going On Behind the Scenes?

We’ve reached out to OpenAI to better understand what’s happening here. Is screaming part of ChatGPT’s training? Or are these anomalies side effects of pushing generative AI to its emotional edge?

While OpenAI has not yet confirmed the cause, their internal risk reports already highlight voice mode as an area needing improvement. Some glitches stem from the AI’s attempt to replicate emotional cues, while others may be unintended artifacts of deep learning.

A Ghost in the Machine?

With more people testing the limits of AI, it’s no surprise that voice mode is being pushed into strange territory. These incidents could indicate bugs or perhaps something deeper, like limitations in how well AI understands (or mimics) human behavior.

So, is this just a funny glitch, or a sign of something more concerning? Either way, ChatGPT’s unexpected screams remind us that even advanced AI can surprise us in truly bizarre ways.

Final Thoughts

OpenAI has made great strides in human-AI interaction, but moments like this reveal how fragile that illusion of control really is. Whether this is just a bug or a breakthrough in realism, one thing is clear:

When your chatbot starts screaming back at you, it might be time to log off.

#ChatGPT, #AIglitch, #GPT4o, #VoiceMode, #OpenAI, #AIScream, #TechNews, #BlackMirrorMoment, #ArtificialIntelligence, #AIHorror