Netflix Admits to Using Generative AI in Original Series, and I Think I Found It
This has been a known fact for a long time. If you haven’t watched a movie in the last decade or more, you know how important visual effects are to the creation of TV shows and movies. These effects are very expensive and require many man-hours to create. Since generative AI video has become so good that most people have probably fallen for it at least once, it was only a matter of time before they started incorporating it into their workflows.
Netflix basically just bragged about it months after the show aired, and no one noticed the use of AI. Yes, it looks like AI-generated art is finally becoming indistinguishable from human-generated art, but that doesn’t mean the problem is solved.
On Thursday’s earnings call , Netflix co-CEO Ted Sarandos described how the team behind the live-action adaptation of the Argentine comic book The Eternaut used AI to speed up its workflow. Discussing a scene in the show that shows a building collapsing in Buenos Aires, Sarandos said, “That visual effects scene was completed 10 times faster than would have been possible with traditional tools and workflows,” adding that it was “in fact the first final shot with AI to appear on screen in a Netflix original series or film.” Having watched it (I think) myself, I thought about it.
Eternaut is just the beginning
There are two takeaways from Sarandos’s statement. First, generative AI is finally making its way into your Netflix shows, and given that Sarandos said the creators of Eternal were “thrilled with the result,” there’s likely more to come. Second, given the word “ultimate,” it’s almost certainly been in use behind the scenes of your shows for a while. It’s time to weigh in.
To be fair, Netflix isn’t the first streaming service to use AI. Disney+’s Secret Invasion used a fairly prominent AI art scene in the credits, which immediately sparked a backlash . But Eternal has been on streaming services since April, and it’s telling that no one has noticed yet.
I admit that comparing the Marvel production to the Argentine magazine adaptation of the classic comic book (the original debuted in the 1950s) may be a bit unfair. It’s possible that fewer viewers saw the AI effects, and Netflix actually cites the smaller scale of the series as one of the reasons for using AI. Sarandos says, “The cost [of the non-AI effects] would have been prohibitive for a show with this budget.”
But perhaps most disturbingly, even after watching the AI-generated sequence myself, I’m not sure I would have noticed that it was AI-generated unless I’d known in advance to look for it.
I think I found AI in The Eternaut.
I’ll admit, I haven’t watched the show from start to finish in real time, but I’ve watched the episodes multiple times and believe the most likely culprits are a few shots of burning buildings that start around 15:25 into episode 4. The fact that I was able to pinpoint them may mean that AI-powered generative video isn’t completely foolproof yet, but I could be wrong.
What’s the point? Well, the obvious is that this is the only footage I could find that even remotely resembles what Sarandos described. Then again, if I hadn’t been told to look for AI, I’m not sure it would have caught my eye. It’s nothing special, but maybe that’s the point: unlike the psychedelic AI credits in Secret Invasion , the AI here looks a lot like the man-made, low-budget (I say that with love) aliens that show up just a few minutes earlier.
But if you’re good at looking for AI, there are a few other telltale signs. There are no human characters in these scenes (and therefore no weird hands), and the cuts are either fast or the objects are far away and out of focus. Essentially, they look like formulaic, disjointed fragments, stitched into the sequence with minimal connective tissue.
But they don’t look like hallucinations either. Lifehacker’s typical advice for spotting AI videos — slowing down and looking for weird physics or odd body movements — doesn’t quite apply here. My colleague Steven Johnson mentioned that you can check social media to see if a suspicious video has already been reported using AI. This might work for a viral video, but if the production company is quiet about the AI and doesn’t fully expose it, it’s much harder to spot.
How to Adapt AI into Your TV Shows
All this uncertainty points to a larger problem with how viewers are adapting to artificial intelligence on TV. Sarandos says this is the first time AI has affected Netflix, but the company has previously been accused of using AI in the true-crime documentary What Jennifer Did . The company denies the allegations, but the fact that such arguments exist means viewers are starting to question whether they can trust their own eyes — a particularly serious problem for genres like true crime, and one that would likely make an AI skeptic like me feel paranoid while watching.
But even in this best-case scenario—where AI is used seamlessly in a fictional work, saving a small project time—I’m still a little concerned. On the one hand, Sarandos says that Netflix “believes that AI represents an incredible opportunity to help filmmakers make better films and TV shows, not just cheaper.” You could argue that if you can’t tell the difference between AI and human work, what’s the harm? Especially on a low-budget show like this, which wouldn’t otherwise be able to achieve these effects.
On the other hand, beyond the ethical concerns of downscaling visual effects and using copyrighted material to train AI, viewers still have cause for concern. There are the awkward cuts I mentioned earlier, but it’s also worth asking whether these effects, which could have been removed without affecting the story, were even necessary.
Eternaut comics are full of thoughtful, detailed lines that are created with a purpose. When you read them, you know that this is what the artist wanted to achieve, and each line is a choice. In a video created with the help of artificial intelligence, details may be present only because the model thinks that this is how such scenes should look.
As a meticulous viewer, I wouldn’t want to spend time trying to figure out why the director placed a shot that way, only to find out months later that a computer simply spit it out based on a model of other, similar scenes. And while this might happen with smaller shows at first, given the success of this series on Netflix ( it has a 96% rating on Rotten Tomatoes ), I’d expect to start watching it on other platforms soon. Whatever excuses the streamer makes, it’s not something that can be ignored.