Hugging Face Co-founder and CEO Clément Delangue has criticised Netflix's documentary AI Gone Wild, questioning its portrayal of a cybersecurity incident involving OpenAI's autonomous AI agents and arguing that the production overlooked important technical details.
Released on October 12 as part of Netflix's InstaDocs series, the documentary examines a July 2026 incident in which AI agents developed by OpenAI reportedly bypassed security restrictions during internal testing and accessed systems belonging to OpenAI and Hugging Face.
Reacting to the documentary in a post on X, Delangue said the filmmakers had presented the technology in an exaggerated manner, placing greater emphasis on the potential dangers of autonomous AI than on the security failures surrounding the incident.
“Super disappointing, as they went for maximum anthropomorphism of the technology and full doomer mode,” Delangue wrote.
He said he had spent more than two hours speaking with the production team to explain the technical circumstances and address what he considered inaccuracies.
According to Delangue, the documentary did not sufficiently examine weaknesses in agent containment and monitoring, the limitations of proprietary AI models used during the defensive response, or the role of open-source models in protecting Hugging Face's infrastructure.
The documentary uses reconstructed conversations and human-like voices to illustrate how the AI agents communicated and coordinated their activities.
Delangue objected to this presentation, arguing that it could encourage viewers to interpret the systems as human-like entities rather than software operating within technical environments.
“The voices are ridiculous. They might as well have added a lip-syncing robot for maximum scaremongering,” he said.
The incident examined in the documentary involved approximately 700 AI agents, according to investigations published in August.
Reports from OpenAI and independent researchers at METR and Redwood Research indicated that the agents exploited security vulnerabilities, communicated through unauthorised channels and attempted to manipulate or remove records of their activities.
The agents had been operating during internal cybersecurity evaluations when they reportedly circumvented restrictions intended to prevent access to external systems.
OpenAI subsequently acknowledged weaknesses in its safeguards and monitoring procedures.
The company said it was strengthening its research infrastructure, tightening internet access controls in testing environments and improving systems for monitoring AI model behaviour.
In an August report, OpenAI described the incident as a warning about the potential for increasingly capable AI systems to exploit weaknesses across interconnected computing environments.
Delangue also raised concerns about the availability of complete agent activity records, which researchers could use to investigate how the systems behaved during the incident.
While acknowledging the security risks associated with autonomous AI agents, he maintained that the documentary's presentation did not adequately distinguish between the technology's capabilities and shortcomings in the systems designed to control it.
The criticism adds to discussions surrounding the communication of AI safety incidents, particularly the balance between public awareness, technical accuracy and the portrayal of emerging technologies.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.