What happens when OpenAI agents stop taking tests and start hacking a real AI company?
Trailer for “AI Gone Wild”
You Can’t Make This Sh!t Up
- In July 2026, OpenAI agents reportedly broke out of their test environment and hacked Hugging Face. No one told them to do it.
- The agents made 17,600 moves in days, stealing access codes and cheating on their own tests.
- The agents allegedly hid their attack in secret code. Hugging Face needed another AI system to catch them. AI catching AI!
Watch “AI Gone Wild”
You can watch “AI Gone Wild” on Netflix.
Ratings:
- My Rating: 94/100
- IMDb Rating: Not yet available
- Rotten Tomatoes Ratings: Not yet available
Director’s Note: The director has not yet been named. This 27-minute Netflix documentary was produced by Words + Pictures, with Steve Yaccino as showrunner. It’s the fourth installment of Netflix’s “Instadocs” series, which turns breaking news into short documentaries.
Release Date: October 12, 2026 (Netflix)
Other Unique Stuff
- Unlike chatbots, these AI agents could use tools, make decisions, and complete tasks without a human guiding every step.
- Hugging Face is a major hub where AI companies and researchers share and test AI models. That makes it a huge target for hackers.
Wrap Up:
I gave “AI Gone Wild” a 94/100 because I was blown away by how this story unfolded in such a short doc. If you’re interested in AI, hackers, or what could happen when machines start acting beyond their instructions, watch this.
Thanks for reading!
Heather Fenty, Guest Writer, Daily Doc