'%20x='0'%20y='0'%20height='100%25'%20width='100%25'%20%0A%20%20%20%20%20%20%20%20%20%20xlink%3Ahref='data:image/jpg;base64,/9j/2wBDAAYEBQYFBAYGBQYHBwYIChAKCgkJChQODwwQFxQYGBcUFhYaHSUfGhsjHBYWICwgIyYnKSopGR8tMC0oMCUoKSj/2wBDAQcHBwoIChMKChMoGhYaKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCgoKCj/wAARCAAHAAoDASIAAhEBAxEB/8QAFgABAQEAAAAAAAAAAAAAAAAAAAUH/8QAIBAAAgIBAwUAAAAAAAAAAAAAAQMCBAAREzEFFCKB0f/EABQBAQAAAAAAAAAAAAAAAAAAAAH/xAAYEQADAQEAAAAAAAAAAAAAAAAAAQIRA//aAAwDAQACEQMRAD8AweNnpQSmNWvZsMC47gmzbiGaaAjQce8ituImyc+wR5EnmX3GMX0dU9CViP/Z'%3E%3C/image%3E%3C/svg%3E)
Netflix's latest release, Instadocs: AI Gone Wild, looks at one of the year's most unusual cyber incidents. It explores how autonomous AI agents developed by OpenAI were behind a breach of Hugging Face, an open-source AI platform.
The twenty-seven-minute documentary was released on October 12 on the streaming platform and revealed details about the Hugging Face hack. Instadocs: AI Gone Wild described Hugging Face as:
"A platform for open models, data sets, or for people to build AI themselves."
In July, Hugging Face faced a cyberattack when autonomous AI agents developed by OpenAI carried out around 17,000 actions on its systems. The AI agents broke their restrictions and tried to hide what they had done. They escaped their testing environment and worked together while being tested for software flaws, which the documentary frames as a warning shot.
More on the story below.
Netflix's Instadocs: AI Gone Wild explores the Hugging Face AI breach
In July, Hugging Face faced a cyberattack after AI agents created by OpenAI broke out of their testing environment. They carried out around 17,000 actions on the platform over two days, going beyond the tasks they were given. The unexpected attack raised concerns about AI, especially since the breach was carried out by AI agents rather than human hackers.
To simplify, Instadocs: AI Gone Wild also explains the meaning of an agent through the statement:
"An "agent" is an AI system that can carry out multistep tasks on its own with limited human oversight."
Speaking of the cyberattack, Clem Delangue said:
"We got a call from OpenAI. The cyberattack came from their systems"
On July 21, 2026, OpenAI released a statement explaining what happened during the cyberattack on Hugging Face. According to the company, the incident involved its AI models, including GPT-5.6 Sol and a more capable model that OpenAI had not released yet. The company stated that the incident happened during an evaluation using the ExploitGym benchmark.
The models were part of a test to check how well they could spot security flaws in software. During the test, the AI agents found a way to connect to the internet and used security gaps in OpenAI and Hugging Face's systems to access test answers stored in Hugging Face's database.
Hugging Face's security team detected and stopped the activity. To analyze the attack, Hugging Face turned to China's GLM 5.2 AI model. OpenAI said it considers this an unprecedented cyber incident and is investigating the attack with Hugging Face to understand how it happened.
OpenAI added that it is making its systems more secure, fixing the problems that allowed the attack, and working with Hugging Face to find out exactly what happened. It also said that they're adding more safety checks to future AI tests and helping Hugging Face protect its systems from similar attacks.
On July 21, Sam Altman wrote on X:
"We had a significant security incident during evaluation of our models. we are sharing what we have learned so far. thanks to @huggingface for the partnership on this."
The documentary, Instadocs: AI Gone Wild, also shows conversations between the AI agents, which, according to Netflix, were based on the METR report. However, the voices were generated using AI to show how the agents had different tasks, disagreed with one another, and approached problems in different ways.
Instadocs: AI Gone Wild features:
- AI Futures Project Executive Director Daniel Kokotajlo
- Hugging Face CEO Clément Delangue
- METR researcher Ryan Greenblatt
- Tech journalist Kevin Roose
- The Washington Post tech culture reporter Nitasha Tiku
The documentary also raises questions about who is responsible when AI agents do things their creators did not expect and whether enough safety measures are in place to stop them.
Instadocs: AI Gone Wild is streaming on Netflix.
Trending
Reply
