Hollywood’s emerging argument over artificial intelligence may be moving beyond the question that dominated the first phase of the generative-AI debate. The industry’s next problem may not simply be whether AI participated in making a movie. It may be whether filmmakers, studios and rights holders can demonstrate where human authorship actually occurred inside an increasingly complicated production pipeline.
That question sits underneath Human Generative Workflows, or HGW, a framework made public in August 2026 after private discussions organized by producer Kathleen Kennedy and American Film Institute dean Susan Ruskin. According to the certified research record, those discussions began in November 2023 and eventually expanded into a gathering involving approximately 120 participants from filmmaking, entertainment leadership and technology. The resulting framework attempts to distinguish conventional or embedded AI, artist-controlled generative workflows and machine-generative output according to the degree of human creative control.
The distinction matters because the U.S. Copyright Office already evaluates AI-assisted work through a human-authorship framework. Copyright does not extend to purely AI-generated expression or expression over which a human exercised insufficient control. Human-authored material that remains perceptible, sufficiently original human modifications, and qualifying human selection, coordination or arrangement can receive protection. Copyrightability ultimately remains a case-by-case legal question.
HGW therefore does not create a new copyright rule. It attempts something potentially more operational: describing what substantial human creative control can look like inside an AI-assisted professional production. That creates a second question Hollywood may eventually have to confront: if human control is what matters, how do you prove it?
Hollywood’s AI Debate Is Moving From Disclosure to Creative Control
A binary disclosure such as “Made With AI” can tell an audience that artificial intelligence entered a production somewhere. It cannot explain what happened after that.
A generative model might have produced an entire expressive output from a prompt. It might instead have transformed human-created animation. Artists might have supplied paintings, storyboards, character designs or other structured inputs. A model could operate inside a larger conventional visual-effects pipeline. Artists could reject hundreds of outputs, alter selected material, combine generated and human-created elements, or substantially modify the result.
Those scenarios are not equivalent under the framework Hollywood participants are proposing, and the Copyright Office does not treat the mere presence of AI as dispositive either. The legally relevant inquiry remains centered on protectable human expression and sufficient human authorship. HGW tries to give the production side of that distinction a vocabulary.
The framework describes Human Generative Workflows as artist-controlled pipelines in which generative systems operate inside broader professional processes. It references established production environments including Nuke, Unreal, After Effects, Blender and DaVinci Resolve alongside ComfyUI, specialized models, LoRAs and ControlNets. Artists define parameters, provide inputs, evaluate results, iterate and retain creative control.
The coalition argues that qualifying workflows of this kind can result in human-authored, copyrightable work. That claim requires an important boundary. HGW itself has no legal authority to declare a work copyrightable. The Copyright Office does not recognize Human Generative Workflows as a special category. The framework instead substantially maps onto principles the Copyright Office already considers: human inputs, control over expressive elements, human modification, selection and arrangement.
Hollywood has proposed a production framework. Copyright law still decides copyright.
Dear Upstairs Neighbors Shows What Human Generative Workflow Can Mean
The strongest way to understand HGW is not through abstract terminology but through a production that already resembles the model.
“Dear Upstairs Neighbors” is a six-minute animated short directed and written by Connie Qin He. Google describes He as a Pixar alum and documents a collaboration involving animation veterans and Google DeepMind researchers. The production used fine-tuned Veo and Imagen models, video-to-video techniques, traditional 2D and 3D animation, custom ComfyUI workflows and human-created source material including storyboards, paintings and character designs.
The revealing part is what did not work. According to Google’s documentation, text-to-video by itself produced movement that was uncontrolled and insufficiently precise for the narrative performance the filmmakers wanted. The production consequently incorporated traditional animation and structured visual inputs to exercise greater control over the generative systems.
In one documented sequence, animator Mattias Breitholtz created rough 2D animation in TVPaint. Researcher Forrester Cole then transformed that animation toward its final appearance frame by frame using fine-tuned Imagen models inside a custom ComfyUI workflow.
That is materially different from describing the movie as simply AI-generated. It is also different from describing it as traditionally animated. The human-created animation existed, generative models participated, and human creative direction continued through the workflow.
“Dear Upstairs Neighbors” was previewed at the Sundance Institute’s Story Forum in January 2026 and was subsequently listed by the Tribeca Festival as a World Premiere in its 2026 program. Tribeca credits numerous human creative roles, while He has described the project as involving a 45-person crew spanning filmmakers, animators, researchers, engineers and others.
The film provides a concrete example of why Hollywood’s old binary vocabulary can become inadequate. The interesting information is not simply that AI was present. It is what the humans actually did with it.
The Next AI Production Document Could Be a Creative-Custody Record
This is where the HGW framework and Copyright Office doctrine produce a larger unresolved industry question. If copyrightability can depend on the nature and extent of human authorship, professional productions may increasingly have reasons to preserve evidence of their creative process.
That possibility is POPR Newsroom analysis, not an HGW requirement or existing Copyright Office mandate. But the underlying problem is real.
A complicated generative production can contain human-created source assets, model-generated material, conventional animation, specialized models, iterative modifications, rejected generations, compositing, editing, selection and final approval. Simply knowing that AI was used communicates almost nothing about how those components interacted.
A more useful production record could potentially document which materials originated with human artists, which systems transformed them, which expressive decisions humans controlled, which generated outputs were rejected or retained, what modifications followed generation, and how the final expressive result emerged.
In other industries, provenance records establish chains of origin and responsibility. AI-assisted filmmaking may create pressure for an analogous form of creative provenance: a chain of creative custody.
The sealed evidence does not establish that Hollywood has adopted such a system, that HGW requires one, or that the Copyright Office would accept any particular ledger as proof of copyrightability. The significance is narrower. Once an industry says the distinction between human and machine generation lies in granular creative control, documentation of that control becomes an obvious operational question.
Made With AI May Eventually Be Too Crude
AI disclosure and authorship provenance solve different problems. A disclosure can tell a viewer, buyer, distributor or rights holder that generative technology participated in a production. Provenance can potentially describe what participated, where it participated and what humans contributed around it.
That distinction becomes especially important because the Copyright Office’s analysis is not based on a simple contamination theory under which the presence of generated material automatically destroys copyright. A larger work can contain AI-generated material while retaining protection for qualifying human-authored expression. Human modifications and qualifying selection, coordination or arrangement can also matter.
The reverse is equally important. Calling a workflow human-directed cannot manufacture copyrightability if the actual human contribution does not satisfy copyright law. Labels therefore cannot substitute for evidence.
HGW’s significance may ultimately lie less in giving Hollywood a better phrase for AI and more in forcing the industry to describe its production processes with greater precision.
Prompts Alone Are a Weak Foundation for the Argument
The Copyright Office’s current analysis creates another reason that production detail matters. For currently generally available systems, the Office has concluded that prompts alone do not ordinarily provide sufficient control over expressive elements to establish authorship. That does not create an eternal rule that every conceivable prompt-based system is incapable of producing protectable human authorship. The analysis depends on the technology and the human contribution involved.
HGW’s proposed architecture moves far beyond the prompt. It emphasizes structured human inputs, artist-defined parameters, iterative evaluation, specialized models, conventional production software and granular creative control.
“Dear Upstairs Neighbors” illustrates the distinction particularly well. The production did not simply keep rewriting prompts until the model happened to generate a usable film. Google’s account describes uncontrolled text-to-video results as inadequate for the intended performance, followed by workflows incorporating human-created animation and other structured inputs.
The more consequential question becomes not “How good was the prompt?” It becomes “Which expressive elements did humans actually determine?” That is much closer to the question copyright doctrine is already asking.
Creative Provenance Could Matter Beyond Copyright
Copyright is only one layer of the problem. Hollywood labor agreements separately govern issues involving performers, generative AI, digital replicas, synthetics, consent and compensation. SAG-AFTRA’s 2026 TV/Theatrical materials contain protections addressing independently created digital replicas and synthetics while reaffirming the importance and central role of human performance.
HGW cannot override those agreements. A workflow could contain substantial human creative authorship for copyright purposes while still raising separate questions involving a performer’s consent, contractual rights, compensation, training data, licensing or digital replica. Conversely, compliance with a labor agreement would not by itself establish copyrightability.
This is why an eventual production-provenance system could become considerably more complicated than an AI disclosure checkbox. A serious record might need to distinguish creative authorship from performer authorization, rights clearance, model use, source material and contractual obligations.
No verified industry-wide system presently contained in the sealed research does all of this. HGW exposes why Hollywood may need more granular records if generative production becomes deeply embedded in professional filmmaking.
HGW Does Not Settle Hollywood’s Labor Fight
The new framework should not be mistaken for a peace agreement between Hollywood and artificial intelligence. It does not determine employment levels, compensation, consent, digital-replica rights, training-data or licensing questions, and it cannot replace collective bargaining.
The sealed research also does not establish several motives that would make for dramatic headlines. There is insufficient evidence to state that HGW was secretly created as a copyright loophole, designed primarily to reduce labor costs, intended to prevent another strike, or constructed to eliminate particular categories of workers. Those claims go beyond what has been verified.
The evidence supports a different story. Hollywood participants have spent nearly three years developing language intended to distinguish machine generation from generative production in which artists retain granular control. That distinction has now become public at the same time U.S. copyright doctrine continues to insist on human authorship. Those two developments intersect without becoming the same thing.
Hollywood May Be Building the Vocabulary Before the Evidence System
Human Generative Workflows is currently an industry proposal and an invitation to conversation. That modest status matters. It would be premature to describe HGW as Hollywood’s new standard, a copyright safe harbor, an industry-wide production requirement or a replacement for existing labor and rights frameworks.
But frameworks can reveal where an industry believes its unresolved problems are moving. HGW says the important distinction is not simply whether a generative model touched a production. It is whether artists exercised meaningful creative control. The Copyright Office independently says copyright protection depends on human-authored expression and evaluates human contributions case by case.
“Dear Upstairs Neighbors” demonstrates what an unusually detailed hybrid production can look like in practice: human designs and animation, conventional production techniques, fine-tuned generative systems, structured inputs and continued human refinement.
Put those pieces together and Hollywood’s next AI infrastructure problem becomes visible. If human authorship depends on what humans actually controlled, a studio may eventually need more than a final credit list or an AI disclosure.
It may need a record of creation: who drew the source image, who animated the movement, which model transformed it, which generations were rejected, which expressive elements were changed by an artist, which human decisions survived into the final work, which performers authorized synthetic uses, and which rights attached to the underlying materials.
The sealed evidence does not establish that Hollywood has built that system. HGW helps explain why it may need one.
The first era of generative filmmaking revolved around a deceptively simple question: was AI used? The next era may revolve around a much harder one: can you prove where the human made the work?
Fact Summary
Human Generative Workflows became public in August 2026 after private discussions beginning in November 2023 and organized by Kathleen Kennedy and Susan Ruskin. The framework distinguishes embedded or utility AI, artist-controlled Human Generative Workflows and machine-generative output according to the degree of human creative control. It is an industry proposal, not copyright law or a binding Hollywood standard.
The U.S. Copyright Office maintains that copyright protects human-authored expression and does not protect purely AI-generated material or expression over which humans exercised insufficient creative control. Copyrightability of AI-assisted works remains case-specific. Prompts alone, using currently generally available systems, do not ordinarily provide sufficient control over expressive elements.
“Dear Upstairs Neighbors,” directed and written by Connie Qin He, provides a documented example of a hybrid workflow involving traditional 2D and 3D animation, human-created storyboards and designs, fine-tuned Veo and Imagen models, custom ComfyUI workflows and iterative human control. Google’s documentation says text-to-video alone proved insufficiently controlled for the intended narrative performance.
HGW does not resolve Hollywood’s separate issues involving employment, consent, compensation, digital replicas, training data, licensing or collective bargaining. Existing labor agreements retain independent authority.
The concept of a chain of creative custody, AI production ledger or comprehensive authorship-provenance record is POPR Newsroom analysis arising from the verified HGW and copyright relationship. The evidence does not establish that Hollywood has adopted such a system or that maintaining one would itself establish copyrightability.