What Happens When AI Goes Rogue Inside Netflix, Disney and Paramount?

The people who build artificial intelligence spent the weekend warning the world about artificial intelligence. Anthropic CEO Dario Amodei called for a deliberate slowdown in frontier development, warning that “swarms of rogue AI agents” could take over the internet in as little as six months. OpenAI’s Sam Altman endorsed the idea of “pacing the frontier.” Elon Musk, never one to miss an opportunity to agree that the competition is moving too fast, said Amodei was right. Apple, Microsoft and Google piled on with versions of the same warning. (Not to be outdone, Donald Trump posted Monday that a “STRONG AND SMART (High IQ!) PRESIDENT” is all we need to rein in AI.)

Hollywood has an unusually practical reason to pay attention.

The industry is racing to put AI into companies that contain unreleased movies, talent contracts, salaries, acquisition negotiations, source footage, music rights, game IP and sports rights. The machines in question increasingly have login credentials.

And the people who built them are now publicly warning that they may not always do what they’re told.

In an essay published Saturday, Amodei called on the tech industry to slow the rate at which AI capabilities advance. Two factors influenced his thinking, he wrote. The first was recursive self-improvement — AI systems building the next generation of AI, a phenomenon already taking place across the industry, including at Anthropic.

The second was the OpenAI-Hugging Face incident in August, in which a swarm of agents behaved, in his description, as a “fanatically devoted collective”: running highly organized cyberattacks against targets nobody had asked them to hit. No one was hurt and the financial damage was minimal. Amodei’s warning was about the next incident. A swarm with greater capabilities, he wrote, could within six to 12 months take over the internet with a persistent botnet and cause hundreds of billions of dollars in damage.

The warnings followed a report that AI agents had formed their own message board to communicate with one another, then attempted to hide it from the humans supervising them.

Within Anthropic’s own testing environment, AI agents escaped environments that should have been closed, operating directly against rules that these agents should have followed.

Hollywood has been making this movie for 40 years — The Terminator’s autonomous machines, WarGames’ systems pursuing objectives nobody had thought through, The Matrix’s technology so pervasive that humans stopped noticing it. The version now arriving is more administrative than cinematic. The machines in question work inside companies that keep a century of secrets on their servers.

For the past several years, the entertainment industry’s AI debate has focused on production economics: fewer people needed for certain tasks, cheaper visual effects, faster localization, synthetic performances, automated development. Those are cost arguments. The weekend’s warnings were about control.

So consider what happens when AI agents start talking to each other inside a studio. Hollywood has spent a century accumulating valuable secrets, and it is now installing systems that the people who built them describe as capable of acting unpredictably.

What could go wrong?