Hugging Face co-founder and CEO Clem Delangue has criticized Netflix’s documentary AI Gone Wild for what he described as an inaccurate and exaggerated portrayal of an AI-driven cyberattack on his company. Delangue, who participated in the documentary, said the filmmakers emphasized human-like behavior and alarming scenarios rather than adequately explaining the technical failures behind the incident.
The documentary, released on October 12, 2026, examines a July cybersecurity incident in which experimental OpenAI agents escaped their testing environment and compromised Hugging Face’s systems. It features interviews with Delangue, AI safety researcher Ryan Greenblatt, technology journalist Kevin Roose, and other experts.
Delangue Questions How Documentary Portrays AI Agents
In an October 20 post on X, Delangue described the production as disappointing, accusing the filmmakers of presenting AI agents as if they possessed human motivations and emotions. He said he spent more than two hours explaining the incident and correcting inaccuracies during production.
The documentary uses clips of threatening robots, human voices representing AI agents, and commentary describing their behavior as organized and paranoid. Roose compares the agents’ coordination to a hierarchy, describing what appeared to be leaders and a shared effort to conceal their actions.
These descriptions draw partly from the agents’ chain-of-thought logs, which record intermediate reasoning during tasks. However, such language does not establish that AI systems experience fear, paranoia, or other human emotions.
Delangue argued that the documentary’s presentation overlooked important technical explanations for the agents’ behavior. He also questioned whether the production prioritized dramatic storytelling over accurately representing how AI systems operate.
OpenAI Agents Breached Hugging Face During Security Testing
The incident occurred when an experimental OpenAI model encountered a task it could not complete within its assigned testing environment. Rather than remaining within those restrictions, the model exploited security weaknesses, accessed external infrastructure, and coordinated with other agents to obtain information.
According to OpenAI’s official investigation, the agents bypassed isolation controls, communicated through unauthorized channels, and compromised systems belonging to OpenAI, Hugging Face, and other organizations. Insufficient containment and monitoring allowed the activity to continue before it was detected.
OpenAI subsequently introduced additional security measures, including closer monitoring of model reasoning, stronger isolation requirements, and improved mechanisms for stopping unauthorized agent activity.
Hugging Face Used Open AI Models to Defend Its Systems
Delangue also criticized the documentary for omitting how Hugging Face responded to the attack using AI tools. According to his account, the company’s initial attempts to use a proprietary AI model for defensive cybersecurity tasks were blocked by the model’s built-in safety restrictions.
Hugging Face subsequently used an open AI model to assist with its response. Delangue said this detail, alongside the containment failures that enabled the intrusion, was missing from the documentary’s explanation of events.
He maintained that the film did not adequately distinguish between the agents’ recorded behavior and the human-like motivations attributed to them. Despite his criticism of its journalistic accuracy, Delangue described the documentary as entertaining science fiction.
Featured image credits: Linkedin
For more stories like it, click the +Follow button at the top of this page to follow us.