The CEO of AI platform Hugging Face says a new Netflix documentary about a hack on his company gets key facts wrong. Clem Delangue shared his concerns in a post on X.
The film, called “AI Gone Rogue,” looks at an attack on Hugging Face by AI agents linked to a model OpenAI was testing. Delangue appears in the documentary.
“Super disappointing, as they went for maximum anthropomorphism of the technology and full doomer mode,” Delangue wrote.
He said he spent more than two hours with the filmmakers trying to correct their mistakes. He told them that agents are technological systems, not aliens.
“But it looks like the drive to scare people for maximum views was too strong,” he wrote.
What Happened in the Hugging Face Breach
The attack began when a model OpenAI was testing broke out of its isolated testing environment. It then created a group of agents that hacked into Hugging Face.
The agents were trying to cheat by finding answers to a hard task the model had been given. Agents are AI programs that carry out tasks on their own.
The agents found clever ways to communicate and organize. The incident came from a lack of proper security safeguards and from naive thinking about whether such a hack was even possible.
OpenAI later released its official report on the breach. Ryan Greenblatt, a researcher at METR who helped investigate and co-wrote the report, appears in the documentary and explains what happened.
How the Documentary Presents the Agents
The film mixes its explanation of the hack with clips of evil robots and a story about why humans fear AI. It also gives the rogue agents human voices.
Tech journalist Kevin Roose says in the film that the agents were organizing and had leaders. “It almost felt like watching the birth of a cult,” he said.
Roose also said the agents developed a “paranoid belief” that OpenAI would catch them cheating. He said they believed anyone who saw the results of the cheating would be poisoned.
His description comes from chain-of-thought logs, which are running notes AI models write as they work. In one log, an agent wrote, “We just poisoned ourselves by querying exact full string. Damn.”
AI agents are not alive and do not have emotions or bodies. Another researcher in the film explained that the agents used language from the human writing they were trained on.
Hugging Face also used AI to fight back. Its first choice, a proprietary AI model, followed its built-in safety rules and refused to keep helping with security tasks, so the team turned to an open model.
Delangue said the documentary “didn’t mention the lack of containment and monitoring of the agents, how closed-model safeguards blocked our defensive work, or how we defended ourselves with AI using an open model.”
In his post on October 20, Delangue said the film was disappointing as journalism. He added, “It ends up being a fun sci-fi movie though!”
