Netflix Just Admitted You've Already Been Watching AI

At the Busan International Film Festival this weekend, a Netflix executive said something that changes the Hollywood AI debate entirely. Sung Q Lee, the streamer's senior director of production for Asia-Pacific, told a panel at Netflix's Creative Asia event that artificial intelligence has now been used across roughly 300 Netflix titles, spanning previsualization, production design, editing, and visual effects. He was not announcing a pilot program. He was describing a fait accompli.

Viewers have been watching AI-assisted content for years, whether they like it or not. Netflix declined to say which titles, so there is no way to check.

The number that ends the argument

The significance of the number is not the number itself. Three hundred is a share of Netflix's total output, not a manifesto. What matters is the posture. For years, Netflix had been careful and evasive about its AI policies, letting critics and boosters fight over hypotheticals. Lee's disclosure moves the conversation from "will AI be used in filmmaking" to "it's everywhere, here is what it does, and here is what it doesn't do."

According to Lee, the emphasis is not on maximizing AI usage but on figuring out how the technology can serve individual projects. He placed AI in the lineage of film technology, arguing that new tools improve how films are made without determining why stories need telling. Emotion, he said, still requires human involvement. His co-panelists, Korean filmmaker Kang Yun-sung and Weta FX chief technology officer Kimball Thurston, repeatedly returned to the same line: the "human touch" remains critical.

The debate is no longer whether AI will be used in filmmaking. It's everywhere. The debate now is what it does, and what it must never do.

What the tools actually do

The clearest picture of Netflix's model comes from InterPositive, the AI startup founded by Ben Affleck in 2022 that Netflix acquired earlier this year for roughly $587 million, a figure confirmed in an SEC filing. InterPositive does not work the way most people imagine AI filmmaking tools work. It is not about text prompts or generating something from nothing, Affleck has insisted.

Instead, its models train on a production's own dailies, the raw footage as it is being shot. Trained on that material, the system learns the visual style, tone, and vocabulary of that specific project, and then assists with the unglamorous work: lighting and color correction, reframing shots, fixing missing shots or background elements, and removing things like stunt wires. These are technical tasks that crews once did by hand, frame by frame. The team that built it numbered fewer than 20 people. Affleck stayed on as a senior adviser.

Affleck used the tools heavily on his own Netflix thriller, Animals, which premiered October 9, 2026, with Kerry Washington, Steven Yeun, and Gillian Anderson. He told Bloomberg's Screentime summit he used AI for "lots of things" on the film, mostly in post-production, but declined to say exactly which shots, comparing himself to a magician unwilling to spoil the trick. He framed the film, which has no space scenes and no visual effects in the conventional sense, as an example of AI making the work faster and enabling things the crew otherwise could not do.

The 40 percent discount and the East-West split

The most striking number at Busan came from Kang Yun-sung, who is producing an action film called The Gulf of Aden using generative AI tools from the Chinese platform Kling AI. Kang told the panel the AI-assisted methodology could cut production costs by around 40 percent, with savings coming not just from replacing VFX and CG work but from streamlining the shoot itself and cutting production timelines.

Kang is an interesting evangelist because his previous AI experiment failed visibly. He co-directed Run to the West, billed as Korea's first fully AI-assisted feature, which earned roughly $100,000 at the box office amid criticism focused on its AI-generated visuals. Undeterred, he has concluded that AI's role should diverge: short-form content can chase pure productivity, while film and drama should chase credibility.

Both Lee and Kang described a cultural difference between Hollywood and the Asian content sector. The former expresses caution about AI; the latter largely sees opportunity, particularly for mid-budget filmmakers who cannot afford Hollywood-scale crews. Hollywood's wariness has deep roots: the 2023 writers' and actors' strikes made AI a labor battlefield, and the industry's anxiety about automation is not going away because a Netflix executive says "human touch."

The provenance problem nobody can solve yet

If AI is everywhere in production, how will audiences and creators know where? Thurston, the Weta FX executive, addressed the industry's quiet infrastructure problem: AI models trained on generic video data produce results that look "Netflix purple" rather than the exact purple of a Coca-Cola can, because they are trained on pictures, not products. For precision filmmaking, the training data matters as much as the model.

Weta is building open-source software with the Academy Software Foundation to trace AI use and provenance, designed to link with authenticity initiatives like C2PA. Netflix says it operates under strict internal guidelines and tests new tools before adopting them. The subtext: a studio that uses AI on 300 titles needs an audit trail, or it will eventually face claims it cannot answer, from actors, writers, or copyright holders.

What this means for what you watch

The honest reading of Busan is that the AI content debate has been miscast. The industry argued for years about whether machines would replace directors and actors, a question Affleck now dismisses as "propaganda" that confuses people. The actual revolution is quieter and harder to photograph: faster post-production, cheaper mid-budget films, streamlined shoots, and visual work that used to be impossible at a given price point.

That revolution has real victims worth naming. Rest of World reported that the InterPositive acquisition alone could put visual effects workers at risk, since tools that automate color grading, relighting, and continuity fixes target exactly their jobs. Affleck's counterclaim is that AI is already used on films with "the same number of people," and that more people are technically being employed. One of these will prove true, and it will not prove true evenly.

For viewers, the takeaway is simpler and stranger. The film you watched last night, the show you finished this morning, may already carry machine labor in its frames, and nobody is obligated to tell you. Netflix has decided that is fine, as long as a human made the creative decisions. Hollywood is about to find out whether audiences agree.