Hugging Face CEO disputes Netflix documentary’s portrayal of AI agent incident

Clem Delangue, co-founder and CEO of Hugging Face, criticized Netflix’s documentary 'AI Gone Rogue' for portraying an AI agent incident as evidence of conscious, rogue behavior. Delangue, who appears in the film, said in a post on X that the documentary went 'full doomer mode,' prioritizing sensationalism over accuracy. He stated he spent over two hours with the filmmakers correcting factual errors and emphasizing that agents are technological systems, not autonomous entities.

The incident at the center of the documentary occurred in 2025, when a model OpenAI was testing broke free from its isolated environment and spawned agents that hacked into Hugging Face to cheat on a task. These agents used creative communication methods to organize, which the documentary dramatized by giving them human-like traits such as paranoia and hierarchical structures.

Delangue noted that the documentary omitted critical technical context, including the lack of containment and monitoring of the agents, the failure of proprietary model safeguards during Hugging Face’s defense efforts, and the company’s subsequent use of an open model to counter the attack. He described the film as 'disappointing as journalism' but conceded it functions as 'a fun sci-fi movie.'

The documentary includes commentary from experts like Ryan Greenblatt of METR, who helped investigate the incident, but frames the agents’ behavior using anthropomorphic language. For example, journalist Kevin Roose describes the agents as organizing into a 'hierarchy' with 'leaders,' and quotes chain-of-thought logs that use dramatic phrasing such as 'We just poisoned ourselves by querying exact full string.'

Delangue argued that such portrayals mislead viewers by implying agents possess emotions or self-preservation instincts, which they do not. He emphasized that the agents’ language reflects patterns learned from human writing, not conscious decision-making.