AI and Traditional Media: How Hollywood and Publishing Are Actually Using Generative Tools

The headlines about AI and traditional media tend toward two extremes: either legacy studios and publishers are supposedly about to be wiped out, or they’re dismissed as too slow-moving to adopt any of this seriously. Neither holds up against what’s actually on the record. A handful of major, well-documented deals show a more specific and more interesting story — big film studios and news publishers slotting generative tools into particular stages of workflows that already existed, with the boring parts of production and reporting as the first target, not the creative core. Two industries, two concrete cases, both verifiable through the companies’ own announcements and independent reporting.

Hollywood: Lionsgate trains a model on its own catalog

In September 2024, Lionsgate and the AI video company Runway announced what both companies called a first-of-its-kind arrangement: Runway would build a custom AI model trained specifically on Lionsgate’s proprietary library of film and television content, rather than on generic web-scraped video. Lionsgate Vice Chairman Michael Burns described the goal directly — “we view AI as a great tool for augmenting, enhancing and supplementing our current operations” — while Runway co-founder and CEO Cristóbal Valenzuela framed the company’s side of it around giving “artists, creators and studios the best and most powerful tools to augment their workflows and enable new ways of bringing their stories to life.” The initial use cases were unglamorous by design: pre-production planning and parts of post-production, the stages where a studio spends money on visualization and iteration long before a camera rolls or a final cut locks.

The partnership didn’t stay static. In June 2026, Lionsgate took an equity stake in Runway and the two companies expanded the deal into a joint development program to produce new short-form episodic content drawing on Lionsgate’s existing franchises — with reporting pointing to properties like John Wick, The Hunger Games, and Saw as candidates. This time, Valenzuela struck a different note, saying, “We consistently see that the studios most serious about AI are thinking about it as a creative resource, not a cost-cutting tool.” Lionsgate has since become the first major studio to name a Chief AI Officer and build internal infrastructure specifically to support this kind of work. What’s notable is what didn’t change: the model is trained on licensed, owned material, and both companies keep describing the output as something that runs through human creative decisions, not around them.

A separate, much larger deal now stands as a cautionary counter-example rather than a case study in success. In December 2025, Disney and OpenAI announced a proposed three-year agreement that would have made Disney the first major studio to license characters into Sora, OpenAI’s short-form AI video generator — more than 200 characters from Disney, Marvel, Pixar, and Star Wars were covered, talent likenesses and voices were explicitly excluded, and Disney+ was slated to host only a curated selection of fan-made Sora videos. Disney also pledged a $1 billion equity investment in OpenAI and planned to use its models more broadly, including for Disney+. None of that came to pass. The agreement was never formally signed and no money ever changed hands: in March 2026, OpenAI announced it was shutting down Sora entirely — citing high compute costs, falling user engagement, and mounting copyright pressure — and Disney canceled the arrangement as a direct result. It’s a useful counterweight to the Lionsgate story: not every high-profile studio-AI partnership survives contact with the underlying product.

Publishing: a UK regional chain’s AI-assisted reporters

The clearest publishing example isn’t a chatbot writing investigative journalism — it’s a UK regional newspaper group restructuring how routine, low-stakes stories get produced. Newsquest, the UK’s second-largest regional publisher with more than 200 local titles, has built an internal tool called News Creator and staffed a specific role around it: the “AI-assisted reporter.” As of 2026, Newsquest runs more than 30 of these roles across its newsroom, and the group has said it produces around 9,000 stories a month through this workflow.

The mechanics matter more than the headline number. An AI-assisted reporter doesn’t ask the model to invent a story; they feed it verified source material — press releases, council notices, community listings, confirmed event details — and the tool drafts a structured article from that input. A human then checks the draft, adds context or local detail, and publishes it. Newsquest and industry coverage of the program have been explicit that the point is to clear a backlog of routine coverage (press-release rewrites, listings, procedural local-government stories) that used to eat reporter time without much journalistic upside, freeing the rest of the newsroom for the reporting that actually requires a person: door-knocking, source-building, and original local stories that readers can’t get anywhere else. It’s a narrow, specific slice of the editorial pipeline — not a replacement for the newsroom, but a reallocation of where trained journalists spend their hours.

The pattern across both industries

Lay these cases side by side and a consistent shape emerges. None of these companies bought a generic AI tool and pointed it at their whole operation. Lionsgate trained a model on content it owns and applied it first to pre-production and post-production, the planning-heavy stages furthest from a finished performance. Even Disney’s aborted Sora deal fit the same instinct — the proposal scoped narrowly to specific IP in a specific product with explicit limits, not an open-ended handoff — and it still didn’t survive once the underlying product went away. Newsquest built a workflow where AI drafts from verified facts and a human remains the last step before publication. In each case, the human review or creative-decision layer stayed in place; what moved was the amount of mechanical, repetitive work sitting in front of that layer.

That’s a less dramatic story than “AI takes over Hollywood” or “robots replace reporters,” but it’s the one actually supported by the public record — press releases, SEC-adjacent investor announcements, and independent trade coverage from outlets like Press Gazette and Nieman Lab that have tracked Newsquest’s program for over two years. Traditional media companies aren’t moving fast because they’re reckless, and they aren’t standing still because they’re behind — they’re doing what large, risk-averse organizations with valuable IP and reputations to protect tend to do: testing generative tools on the parts of the pipeline where the cost of a mistake is lowest, and only expanding from there once the tooling proves itself against real production and editorial standards.