Google, A24, and the Quiet Frontier: Why the Battle Over AI in Film Begins Before the Camera Rolls

The deal everyone is misreading

On June 22, 2026, Google and A24 announced a research partnership, and Google took an equity stake in the studio — reportedly around $75 million, a figure that comes from Wall Street Journal reporting rather than any official Google disclosure. Most of the coverage has framed it as a financing event: Big Tech writes a check, indie studio takes the money, fans revolt. That framing is not wrong. It is just shallow.

Let me establish the lens before the argument, because the lens is the whole point here.

What Google bought is not A24’s library. The deal explicitly walls the catalogue off; in Google DeepMind’s own words, it is “not a production deal, or an IP deal, or a data training deal.” What Google bought is something less tangible and more valuable: cultural legitimacy. A24 is the most trusted brand in independent film — the studio behind Moonlight, Everything Everywhere All at Once, and Anora, a brand people get tattooed on their bodies. If artificial intelligence can be made acceptable there — inside the house that taste built — it becomes acceptable everywhere else by implication.

That is the trade. A24 gets capital and access to AI infrastructure. Google gets a seat in the room where the next Moonlight gets made, and the right to say that the most artist-forward studio in the business builds with them.

The frontier is upstream, not on screen

Here is the reframe I want to offer. The public conversation about AI in film is fixated on the wrong frontier. We are all watching the screen — waiting for the synthetic actor, the deepfaked star, the fully AI-generated feature. We are guarding the exit.

The technology is coming in through a different door entirely.

A24 Labs — the studio’s roughly 20-person technology division, led by Scott Belsky, the co-founder of Behance — has announced that its first tool will help generate storyboards. Not finished footage. Storyboards. Previsualization. The early visual artifacts through which a film is imagined before it is shot.

This matters more than the headline suggests, because pre-production is not a harmless sandbox. Storyboards, concept art, mood boards, and look-books are where a film acquires its visual language. They shape budgets, casting, production design, and investor confidence. And critically, they are paid creative labor — the work of human artists whose contribution is invisible to the audience but structural to the film.

The displacement that arrives first will not be a synthetic Tom Hanks. It will be the quiet automation of the sketches and boards and references that happen before a camera ever rolls. The most controversial AI images in cinema may be ones that never appear on screen.

The same month, the same defense, the same objection

We do not have to speculate about how this lands, because the experiment already ran — twice, in the same month, with the same result.

Twenty days before the Google–A24 announcement, Martin Scorsese — the most decorated living American director, a man synonymous with film preservation and the defense of cinema as art — joined the AI company Black Forest Labs as an adviser. He used its FLUX model to storyboard a scene from his next film. He was careful and specific: pre-visualization only, never final footage, a way to communicate his vision to his human crew faster. He called it “creatively freeing.”

One week later, the Art Directors Guild — IATSE Local 800, which represents the storyboard artists, concept artists, and production designers whose work this is — formally condemned him. Their argument was not that he had replaced his actors. It was that the tool does “the jobs that are rightfully the jurisdiction of Art Directors Guild Local 800 artists and designers.” The concept artist Karla Ortiz put it more bluntly, accusing him of throwing the storyboard artists he had worked with throughout his career “under the bus.”

Notice the structure. Scorsese used the exact defense A24 and Google will use: we only use it for pre-viz; creative control is preserved; no final footage is generated. And that defense was rejected immediately — not by AI opponents in the abstract, but by the specific people whose jurisdiction pre-viz represents. The “storyboard-only” shield has already been publicly stress-tested and found wanting. A24 should expect the same response, because it is building the same tool.

That convergence — two of the most respected names in cinema, the same justification, the same objection from the same constituency, all inside a single month — is the actual news. Not the dollar figure.

What I actually think — and where I depart from both camps

I do not arrive at this as a pessimist about AI. Readers of my work know I believe the opposite of the replacement narrative: the real power of these tools is augmentation — what a person can now do that they never could before. I have argued that there is a third territory between pure-human and pure-AI art: AI-enhanced human art, where the tool does the journeyman’s labor so the artist can pour themselves more fully into the part only they can do. I still believe that. A young filmmaker in San Juan or Dakar who could never afford a previs department, now able to block out a sequence overnight, is a genuine expansion of who gets to make films. That is real, and it is good.

But neither AI-evangelism nor AI-rejection survives close inspection, and the honest position requires holding two things at once.

The deal itself is defensible — arguably the most artist-cautious of the major studio–AI tie-ups to date. Capital flows into the studio rather than the studio surrendering equity for tools. The library is explicitly walled off. The “no training, no IP” structure is precisely designed to avoid the scraped-data accusation that haunts FLUX, whose training sources remain undisclosed. Compare that to Netflix’s acquisition of Ben Affleck’s InterPositive, reportedly tied to a patent describing 10–20% below-the-line cost-cut targets, and A24’s terms look genuinely more respectful.

And yet. Structure at launch is a promise, not a guarantee. The thing you say before the tools become load-bearing is that they will “preserve creative control.” The question that decides everything is one I have asked in every other context where a productivity technology meets human labor: who captures the gain?

Businesses do not exist to create jobs. They exist to create value, and employment is how a society distributes the gains of that value creation. AI is a productivity technology. When it makes storyboarding faster and cheaper, the gain is real — and it will flow either to the workers (better pay, more ambitious work, more films attempted) or exclusively to capital (higher margins, fewer artists hired). Nothing about the technology decides which. That is a design choice, and we are making it right now, deal by deal, with very little deliberate planning.

This is not a reason to ban the tools or to boycott the studio. It is a reason to be specific about what would make this deal genuinely artist-positive rather than extractive: enforceable creative-control commitments rather than rhetorical ones; transparency about what the shipped tools actually do; preserved compensation and credit for the storyboard and concept artists whose tasks the tools touch; and a library wall that holds in practice, not just at the press conference.

The bridge we are not planning

There is a number worth sitting with. In Pew’s Americans and AI 2026 survey, among adults aged 18 to 29 — which is to say, A24’s core audience, the people who get the tattoos — 48% say AI will have a negative impact on society, against just 14% who say positive. That is the brand risk A24 just took on, and it explains why the backlash was instant.

But the deeper point is the one my grandfather’s trade taught my family without anyone intending it to. Disruption does not announce itself through the front door. The blacksmith was not put out of business by a better hammer; he was put out by a machine arriving in an adjacent industry entirely. By the time the change is obvious, the adaptation window has already closed for those who waited.

We are at that moment with AI in film. The change is real, it is already underway, and it is arriving upstream — in the imagined film that exists before the real one is made. The losing strategy is to keep guarding the screen while the frontier moves to the sketchpad. The winning strategy — for studios, for guilds, for the artists themselves — is to decide, deliberately and now, who gets to imagine the film, and who gets paid to do it.

That is not a question Google or A24 can answer for us. It is one the industry has to answer for itself, before the tools become load-bearing and the answer hardens into a default nobody chose.