A24 walked away from the 2023 Academy Awards with seven trophies. Everything Everywhere All at Once was made for $14 million and earned $69 million at the worldwide box office — rewriting the independent studio playbook from production to marketing. A24 greenlights maybe ten projects in a good year. Its brand was built on one thing: a consistent, uncompromising aesthetic eye that could sift through thousands of scripts and know which ones were worth making, and which directors were right for them. So when news broke that Google DeepMind had committed $75 million to co-develop AI filmmaking tools with the studio, the industry's attention landed less on the dollar figure than on the name attached to it. A24 had always seemed like exactly the kind of studio that kept AI at arm's length. The fact that it didn't is what made people lean in.
How you read this deal for the creative industry depends on whether A24 is simply providing training data. Look more closely at what DeepMind actually wanted from this arrangement, and the nature of the contract reads very differently.
What DeepMind Was Really Buying: Editorial Taste
According to TechCrunch's reporting, the two companies agreed to co-build AI-powered filmmaking tools. DeepMind brings the technical capability; A24 brings its production pipeline and its accumulated aesthetic judgment. The $75 million funds that tool development.
A tech company backing a content company is nothing new. But this deal is structurally different from Netflix investing in production studios or Amazon acquiring MGM. If the goal had been distribution reach or IP ownership, a studio acquisition or content licensing agreement would have done the job. What DeepMind chose instead was joint tool development — meaning direct access to A24's production floor and, crucially, to its decision-making process.
There's a reason for that. The more technically capable AI video generation tools become, the harder they run into the bottleneck of aesthetic judgment. Tools like Sora and Runway can produce seconds of video from a text prompt. But determining whether a scene is cinematically alive, whether an editing rhythm creates the right emotional arc, whether a line of dialogue actually builds a character — those are evaluations AI models struggle to make on their own. What A24 has spent two decades accumulating in the independent film ecosystem is precisely that judgment: a deeply organizational sense of which director fits which project, which script can hold an audience in a theater. The number of decisions an editor makes over the course of a single film runs into the thousands. Those decision patterns can generate the feedback signals an AI model needs to be trained on quality, not just output. DeepMind didn't invest in A24's movies. It invested in A24's ability to judge which movies deserve to exist.
The Concern About A24's Independence Is Real
This deal doesn't read as purely positive. Once you understand where A24's competitive edge actually comes from, the degree to which this contract touches that edge becomes hard to ignore.
A24 is a studio that chose aesthetic autonomy over the financial backing of major distributors and streaming platforms. Films like Midsommar and Ex Machina — where artistic vision took precedence over market safety — are what put the studio where it is today. So what happens when an outside funder's needs enter those creative decisions? With DeepMind's $75 million in the picture, there is now a structural risk to editorial independence. The question "Is this scene useful for AI training?" could come into tension with the question "Is this scene cinematically necessary?" Creative partnerships where funder priorities seep into content direction are not rare — they are a repeating pattern.
There are also legitimate concerns around copyright and creator rights. A central issue in the 2023 Writers Guild of America (WGA) strike was AI-related contract language: writers objecting to their scripts being used without consent to train AI models, and to contracts that would effectively require humans to polish AI-generated content. The terms under which the A24-DeepMind deal would data-ize writer and director creative judgment have not been made public. Critics in the film community are already raising concerns that the partnership will dilute A24's curatorial principles over time. That opacity is itself a substantive reason for concern.
What Content Creators Should Be Watching in a World Where Taste Becomes Data
For independent content creators and directors — in Korea and elsewhere — it's worth stepping back and thinking through what this deal actually signals.
First, it's worth mapping where AI is now crossing new lines in creative fields. Video editing assistance, script drafting, image generation are already standard in many workflows. The DeepMind-A24 deal is a signal of the next layer: AI moving beyond production assistance toward systems that can evaluate whether content is good. We already see earlier versions of this — AI predicting YouTube thumbnail click-through rates, forecasting newsletter subject-line open rates. The DeepMind-A24 project is an attempt to apply that same capability to film, a far more complex and long-form format.
The ownership structure of taste data is shifting, and that's worth paying attention to. A creator's editorial instincts and aesthetic judgment have always been personal, tacit knowledge — hard to quantify, impossible to transfer. Once studios begin converting that knowledge into AI training data through contractual arrangements, individual creators have reason to start reading the data usage terms of the AI tools they rely on daily. Adobe Firefly, Canva's AI features, video editing platforms — all of them apply. Many generative AI services include clauses that allow user uploads and editing decisions to be used for model improvement. Knowing where your taste goes is different from not knowing.
Articulating your own judgment criteria is also practical preparation. If you can explain why you used a particular scene, why a certain direction fits your audience, you have a stable reference point for evaluating what AI tools generate versus what you would create. That articulation has a direct practical payoff: more specific prompts produce better results. Just as an A24 director's ability to describe their own cinematic language can become meaningful signal for model training, a solo creator's taste vocabulary shapes the quality of what they get from AI tools.
The speed at which AI tools are restructuring the cost of content production also deserves attention. Tools backed by DeepMind-scale investment are likely to reach commercial availability within three to five years. Post-production, subtitle translation, music licensing — these are precisely the cost centers that those tools are designed to compress. Whether this is an opportunity or a threat depends entirely on what a creator's actual competitive advantage is. For creators competing on cost efficiency, it's a threat. For creators competing on taste and judgment, cheaper production tools are simply good news.
There's an interesting paradox embedded here. The more AI tools raise production efficiency, the more differentiation flows toward people who know what's worth making. If you can generate a thousand images in an hour, your competitive edge lives in the sense to pick ten. Researchers who study the future of work have been pointing toward this for years. As AI absorbs repetitive production, human judgment and taste don't become obsolete — they become a narrower, sharper competitive point.
DeepMind didn't choose A24 for its filmmaking technology. It chose A24 for its ability to judge which films are alive. As studios begin ceding that judgment to AI systems, the most durable advantage left for individual creators is building that same capacity — and keeping it in-house.
