Netflix used generative AI on 300 titles this year, and there’s no rulebook
Follow Us
The latest Netflix shareholder letter casually throws out a number. About 300 movies and TV shows released on the platform this year were made using generative AI technology – mostly in the post-production process. The co-CEO of Netflix, Ted Sarandos, puts this into the category of a cost and time-saving achievement – shots that would have been left out, now produced twice as quickly and at a much lower cost. This sounds like yet another standard studio productivity move. It’s not.
SurveyAlso read: What is Kimi K3: Moonshot’s frontier AI model that matches Claude Opus 4.8 and GPT 5.5
Last year, Netflix’s application of generative AI technology was an experiment that took the form of a single sequence within the science fiction series The Eternaut. Today, its applications include crowd multiplication, reenactments of historical battles and world-building shots within movies and TV shows such as Glory (Indian production), Brasil 70: A Saga do Tri and The American Experiment documentary series, which has no less than 17 minutes of AI-generated content. This is not an incremental use of technology. This is an entire production pipeline rebuilt using new tools over the course of only a year by a company that boasts nearly one billion users and sets industry standards.
At some point along the way, during all of this scale-up, Netflix went beyond a certain threshold that needs more examination than has occurred. Its latest unscripted series “Wonka’s The Golden Ticket” features a voice synthesized by artificial intelligence based on an actor that died in 2016 and did not give any permission for his image or voice to be used in projects posthumously. It does not matter whom exactly was behind creating this voice or where the source audio came from and who approved it. What is much more important than the technology itself is silence around it. Voice cloning of a living actor at least creates the issue of consent and compensation that was recognized by unions.
Also read: What Is Inkling: Thinking Machines Lab’s bet on customisable AI
This is actually the story that gets buried by Netflix’s headline of 300 titles. The point isn’t that Netflix uses generative AI; every big movie studio is on the way to this point. The point is that Netflix has gone from using AI to assist with its visual effects work, adding a few more extras to a crowd or expanding a battle sequence, to using AI to recreate performances, and there is no public framework for drawing the line between those two points. There is no disclosure mechanism to help moviegoers understand when an image has been generated rather than shot. There is no guidance about whether an actor’s voice needs permission like a script needs rights.
Sarandos has always been clear about the Netflix story: AI makes it possible for Netflix to do shots that it can’t afford to do otherwise. This is a true story for production companies, but it completely avoids the harder question. Cost is why AI was used. It doesn’t answer why there isn’t a boundary drawn around where it should not have been used.
Netflix is not breaking any laws here, and perhaps not even doing anything that won’t become common among other streaming services over the next couple of years. This is precisely why Netflix should have never gotten away with it in the first place. After all, the one setting the standards for the norm here tends to be the first one to establish the standard, and what the current standard seems to be is simple: adopt it, and try to explain it if you manage to remember. And there are still three hundred movies out there without an explanation.
Also read: This OpenAI model breached a vending machine to prove a point on AI safety
