AI in Film and TV: Three Legal Battlegrounds Every Creator Needs to Understand

Generative artificial intelligence has moved from experimental novelty to production tool in film and television, and the law has not kept pace. In the United States, three distinct conflicts now define the landscape: whether AI-assisted work can be copyrighted at all, whether training AI models on copyrighted material is legal, and whether creative professionals can protect their compensation, artistic control, and likenesses. The U.S. Copyright Office has staked out a firm position on the first question—human authorship is required—while the second remains unresolved in the courts, and the third is being fought out in collective bargaining rooms rather than courtrooms.

Setting the Stage: AI's Impact on Film and Television

Since launching an initiative in early 2023, the U.S. Copyright Office has been examining the copyright questions raised by artificial intelligence, including the scope of copyright in AI-generated works and the use of copyrighted materials in AI training. After public listening sessions and a notice of inquiry published in the Federal Register in August 2023—which drew more than 10,000 comments by that December—the Office began issuing a multi-part report analyzing the issues. That an agency devoted to registering creative works felt compelled to mount such an inquiry signals how thoroughly AI has unsettled assumptions that governed entertainment for over a century.

The disruption is not hypothetical. Legal analysis of AI in entertainment notes that streaming platforms already use AI to personalize recommendations and that studios employ AI-driven sentiment analysis to tailor promotional strategies. The same analyses identify concrete harms: automation threatens traditional roles for writers, editors, and visual artists, while AI-generated digital replicas that can mimic an actor's voice or likeness without consent have sparked debates over ethical and legal protections. The technology's reach into both the business and the craft of filmmaking is why guilds, studios, and regulators are all moving at once.

The Human Authorship Standard: Copyrighting AI-Generated Content

The central legal boundary is the Copyright Office's position that works are not copyrightable unless sufficient human creativity is involved. The Office has consistently maintained that human authorship is a prerequisite for protection, a position supported by recent court decisions. The underlying principle is that copyright protects the human expression of ideas—not the ideas themselves, and not works generated autonomously by machines.

For filmmakers and showrunners, the practical consequence is that a purely AI-generated scene, image, or script element sits in the public domain: no one owns it, and no one can stop a competitor from copying it. But the Office has clarified that if a human provides significant creative input—such as editing, arranging, or selecting AI-generated elements—a work might be eligible for copyright protection. The extent of human involvement and control becomes the crucial factor. A director who uses AI to generate raw footage and then substantially reworks it stands on much firmer ground than one who prompts a model and ships the output. Legal commentary cautions that this ruling does not fully address the complexities of human-AI collaboration, and the line between human authorship and machine generation is still being refined through registration decisions and litigation.

The Battle Over Training Data: Fair Use and Infringement

The second conflict concerns inputs rather than outputs. Large AI models require vast datasets for training, much of it compiled from publicly available but often copyrighted images and texts. AI companies argue that this scraping constitutes fair use, claiming the resulting models are transformative and do not compete with the original content. Their opponents see uncompensated appropriation. Legal cases involving data scraped from the legal research platform Westlaw and from major publications such as The New York Times illustrate how contested the fair use doctrine has become in the AI context.

The Copyright Office has acknowledged the fair use debate as part of its ongoing examination, but no definitive ruling has been issued, and the lawsuits continue. Policy analysts tracking the Midjourney litigation describe a regulatory gap that has created widespread uncertainty for AI service providers, their users, and creative professionals alike. A second complication compounds the first: AI-generated outputs can closely resemble existing copyrighted works, raising infringement questions even after a model is trained. For creators whose past work may have been swept into training datasets without consent, the unresolved status quo means their material may be generating value for AI developers while they receive nothing.

Protecting Creative Professionals: Guilds' Demands for Guardrails

While the courts weigh copyrightability, the guilds have negotiated their own protections. The Directors Guild of America, representing more than 19,000 directors and members of directorial teams, told the Copyright Office in its AI study comments that "It is essential for policymakers and the courts to strike a balance… by developing meaningful and enforceable guardrails to prevent technological advancements from undermining intellectual properties, job opportunities, and artistic integrity." The DGA emphasized that "The technological boundaries of [generative] AI are virtually limitless, emphasizing the need to safeguard the integrity and the singular creative vision of a director."

The Writers Guild of America translated similar concerns into contract language. Under the 2023 Minimum Basic Agreement, neither traditional AI nor generative AI is considered a writer, so no written material produced by either can be considered "literary material" for compensation or credit purposes. If a company hands a writer AI-generated material that has not been previously published or exploited, that material does not count as assigned material or source material—and a company cannot give a writer an AI-generated screenplay and pay only a rewrite fee; the writer is treated as the first writer of the screenplay.

Companies must also disclose when materials given to a writer were generated by AI, and the WGA reserves the right to assert that exploitation of writers' material to train AI is prohibited by the MBA or other law. Writers may choose to use AI in performing writing services if the company consents and applicable policies are followed.

These guardrails are specific to guild agreements and may not apply universally across all creative roles or productions. The WGA has also pressed its case in Washington: more than 500 members submitted comments in the Copyright Office proceeding, and guild bodies co-signed a March 2024 letter to Senate Majority Leader Chuck Schumer calling for AI protections for film and television writers, journalists, and content creators.

Key Conflict Zones in AI for Entertainment

The three battlegrounds are often conflated, but each turns on a different question, involves different stakeholders, and carries different risks. The table below separates them and contrasts the official legal stance with the positions industry advocacy groups have taken.

The three primary conflict zones in AI for entertainment, with stakeholder positions and implications for creators
Conflict Area	U.S. Copyright Office Position	Creative Guilds' Position	Key Implication for Creators
Copyright of AI-generated works: who owns the intellectual property of content created with AI?	Human authorship is required for protection; purely AI-generated content is not copyrightable, though significant human editing, arranging, or selecting may qualify a work.	AI is not a writer or creator; material produced by AI does not count as "literary material" for compensation or credit under agreements like the WGA MBA.	Creators must ensure significant human creative input to secure copyright for AI-assisted works; the boundary is still being refined.
Use of copyrighted training data: is training AI on existing works fair use or infringement?	The Office is examining the issue and acknowledges the fair use debate; no definitive ruling has been issued.	The WGA reserves the right to assert that exploitation of members' material to train AI is prohibited by contract or law.	Legality remains uncertain, posing risks for AI developers and for creators whose works may be used without consent or compensation.
Protection of creative labor and likenesses: how are jobs, artistic vision, and digital replicas safeguarded?	The Office's initiative includes study of these issues, including digital replicas.	The DGA demands enforceable guardrails protecting intellectual property, job opportunities, and artistic integrity; guilds seek consent requirements for digital replicas.	Protections exist mainly through guild agreements, which may not cover all roles or productions.

The three primary conflict zones in AI for entertainment, with stakeholder positions and implications for creators

Conflict Area

U.S. Copyright Office Position

Creative Guilds' Position

Key Implication for Creators

Human authorship is required for protection; purely AI-generated content is not copyrightable, though significant human editing, arranging, or selecting may qualify a work.

AI is not a writer or creator; material produced by AI does not count as "literary material" for compensation or credit under agreements like the WGA MBA.

Creators must ensure significant human creative input to secure copyright for AI-assisted works; the boundary is still being refined.

Use of copyrighted training data: is training AI on existing works fair use or infringement?

The Office is examining the issue and acknowledges the fair use debate; no definitive ruling has been issued.

The WGA reserves the right to assert that exploitation of members' material to train AI is prohibited by contract or law.

Legality remains uncertain, posing risks for AI developers and for creators whose works may be used without consent or compensation.

Protection of creative labor and likenesses: how are jobs, artistic vision, and digital replicas safeguarded?

The Office's initiative includes study of these issues, including digital replicas.

The DGA demands enforceable guardrails protecting intellectual property, job opportunities, and artistic integrity; guilds seek consent requirements for digital replicas.

Protections exist mainly through guild agreements, which may not cover all roles or productions.

Read together, the table shows why the debate feels so tangled: the copyrightability question has an official answer, the training-data question does not, and the labor question is being answered contract by contract rather than by any single authority.