By: Zach Miller

Last December, a four-part reality show called Non-Player Combat premiered on YouTube. Every frame of it was generated by AI. Every character was synthetic, every camera angle was a render, every line of dialogue was spoken by a voice that did not exist. The whole series, four full-length episodes, was produced in under two months for roughly $28,000, a figure its creators claim is 90 percent less than the cost of a comparable human-made production. The show has real viewers, episode counts, and thumbnails tuned for the YouTube algorithm.

John Chachas, the veteran media investment banker who built his career on the consolidation of American broadcasting, has been watching this kind of content accumulate for the better part of a year. What he sees does not look like a novelty.

“The amount of stuff now on the airwaves, available online, available on these platforms, that’s being produced largely by one person sitting behind a computer using AI tools is astonishingly large,” he said in a recent interview. “Astonishing.”

The Volume Problem Hollywood Is Still Negotiating Around

The guilds that shut down Hollywood for most of 2023 secured meaningful concessions from the studios on the use of AI. The Writers Guild agreement specified that no form of AI, generative or otherwise, can be considered a writer. The SAG-AFTRA contract required explicit consent and compensation each time a performer’s digital replica is used. Both agreements were hard-won. Both operate inside the universe of signatory production companies that employ guild members.

Non-Player Combat is outside that universe. So is Exit Valley, a satirical series set in a simulated Silicon Valley, which runs on Showrunner, a platform backed by Amazon that lets users type a prompt and produce a full episode. The guilds’ jurisdiction ends at the boundary of the signatory contract. A college student with an RTX 4090 and a YouTube channel does not sign a SAG-AFTRA contract. A solo producer operating from a laptop in Ostrava does not owe the WGA a writer’s-room credit.

Chachas has been watching the external pressure show up in his own feed.

“I watched the first ten minutes of one episode of Heated Rivalry on YouTube,” he said, referring to the hockey-romance adaptation that went viral in late 2025. “My YouTube inbox is now literally littered with what I’d bet are AI-produced shorts about two men. That’s the stuff taking root in how we all consume. It’s both powerful and disconcerting to me.”

The algorithmic recommendation surfaces on YouTube, TikTok, and Meta’s reels product does not differentiate between human-made and AI-generated content in its ranking signal. What it rewards is watch-time, completion rate, and engagement density. A synthetic 60-second clip that keeps a viewer engaged scores identically to a human-made one. That is the economic bridge the guilds’ 2023 contracts do not span.

Sora 2 and the Production-Cost Curve

OpenAI released Sora 2 in late September 2025. The model, which adds synchronized dialogue and sound effects to its predecessor’s video capabilities, produced demo content that film-industry executives told IndieWire was qualitatively different from what came before. What the demos actually illustrate is the collapse in the underlying cost curve.

An establishing shot that requires a camera crew, a drone operator, a location permit, and a travel day now costs the output tokens of a large video model. A background actor whose role consists of walking through a frame holding a coffee cup can be rendered, with a fully-licensed digital replica, for the compute cost of a few seconds of inference. Hollywood Reporter testing with Sora beta users found that entire opening sequences for feature films could be generated in an afternoon with a single person at a workstation.

Chachas frames this the way a media banker frames any cost curve.

“In the TV production world, there are now 30- to 60-minute programs being produced completely by AI,” he said. “And it’s pretty high quality. You could figure it out and see it, but it’s close. It’s going to significantly alter the economic model for production by driving down cost. But it’s also going to drive down the value. It’s a little bit like littering the marketplace with all kinds of crap produced on a computer and not produced by a typical team of people.”

A glut of low-cost synthetic content compresses the unit economics of the entire medium at the same time it displaces human labor. When the marginal cost of producing a watchable thirty-minute episode approaches zero, the price a streaming platform or an advertiser will pay for any individual episode follows.

The Copyright Problem Nobody Has Solved

Content created without meaningful human authorship cannot be copyrighted in the United States, per the US Copyright Office’s January 2025 guidance. A studio that generates an establishing shot with Sora cannot register that shot as its own work. A channel that runs a fully AI-generated reality show on YouTube cannot assert copyright claims against anyone who scrapes and re-uploads the footage.

Talent agencies, led by WME, have signaled that they will aggressively defend client likenesses against unauthorized synthetic replication. The legal theory available to them is right of publicity rather than copyright, a state-law patchwork that varies from California’s well-developed framework to near-nonexistent protection in some smaller states. Enforcement against a creator operating from outside the United States is, in practice, difficult.

What Chachas Is Warning About

Chachas’s argument is that the business model that supported television production for seventy years is decomposing faster than the protective frameworks around it. Broadcast television advertising is in structural decline. Cable carriage fees are compressing. Streaming has not replaced either revenue stream; it has replaced the audience behavior. What has replaced both is a recommendation surface that rewards volume, novelty, and engagement density. AI-generated content is optimized for all three.

On the broader side, Chachas has been telling anyone who will listen that the regulatory conversation about AI is happening at the wrong altitude.

“AI presents opportunity and enormous risk, enormous risk,” he said, repeating the phrase for emphasis. “And I don’t think anybody in government is really paying attention. From top to bottom, all they care about is: are they first in the race and making lots of money?”

He returned in the interview to a comparison that has become his shorthand: the 1983 film WarGames, in which a networked defense computer nearly launches a nuclear strike before being talked down.

“We are very close, very close, to the kinds of things where we could look back and go, ‘Wow, someone actually turned the missiles on. Who did that? ”he said. “At some point, you will not be able to control it anymore. That’s pretty scary to me.”

The Near-Term Takeaway

For anyone working in or around the content business, the near-term question has moved past whether AI-generated programming will displace traditional production. A segment of it already has. The operative question is how quickly the displacement scales, which parts of the value chain get compressed first, and how the remaining human-authored content differentiates itself once the synthetic baseline becomes abundant.

“If AI creates this stuff at almost pennies compared to typical production, what happens to the compensation of writers, videographers, showrunners, and, of course, talent, actors?” Chachas said. “Their value will be under enormous pressure.”

The volume Chachas is watching accumulate on YouTube, on Showrunner, and on every other algorithmically sorted surface moves on its own schedule. It is already on the air.