Hollywood is putting artificial intelligence on the legal docket. At Deadline’s inaugural L.A. Law Summit last week, a panel of entertainment attorneys warned that AI is creating both new creative opportunities and unprecedented legal questions—from who owns AI-generated material to whether copyrighted works can be used to train models without permission or payment.
The stakes are enormous. CAA Chief Compliance Officer and Deputy General Counsel Tammy Brandt told the summit that “cultural change happens a long time before political and legal change does,” cautioning that the industry must “lean in” to AI while “fighting for the rights of the content creators and intellectual property.”
As AI tools become embedded in production workflows and AI companies face mounting copyright litigation, the entertainment industry is confronting a fundamental question: who controls AI-generated entertainment, and who is liable when it infringes on someone’s rights?
The Deadline L.A. Law Summit, held in mid-September 2026, brought together lawyers, talent, agents, and producers for a half-day of sessions on intellectual property, M&A, and the creator economy. The closing panel—”AI: Navigating the Realities & Possibilities of Digital Hollywood”—featured Brandt alongside Glaser Weil’s Jesse Levin and Blank Rome’s Craig Weiner.
The discussion reflected an industry that has moved past debating whether AI belongs in Hollywood to grappling with how to regulate it. “AI is here to stay—it’s an incredible technology,” Brandt said. “We have to lean in, but we have to keep fighting for the rights of the content creators and intellectual property. That’s the most important thing.”
Weiner highlighted advertising statutes “across the country” as tools for protecting clients’ images and work, beyond traditional cease-and-desist letters. Levin addressed the challenge of gaining leverage over well-funded AI companies, noting that “they’re the deep pocket, but they’re not the ones out there making all of this content.”
The summit’s timing was notable. It occurred as OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei addressed the United Nations about AI’s risks and rewards—a reminder that Hollywood’s concerns are part of a broader global reckoning.
Despite the legal uncertainty, AI offers tangible benefits across the production pipeline. The technology is already being used in visual effects, pre-production concept development, localization and dubbing, editing assistance, and virtual environment creation.
The cost and time savings can be dramatic. CJ ENM, a major Korean content company, produced the feature film The House using a hybrid AI-live action pipeline. Actors performed against green screens, while all backgrounds and visual effects were AI-generated. The film cost approximately $360,000 and completed principal photography in just four days—a fraction of conventional production costs.
In China, AI-driven production has accelerated dramatically. Data from the National Radio and Television Administration shows that over 90% of the 430,000 micro-dramas launched in the first eight months of 2026 were AI-generated. The AI drama market reached 22 billion yuan in the first five months of 2026, with full-year projections of 40 billion yuan.
These examples illustrate AI’s potential as a tool that expands creative possibilities rather than simply replacing human labor. However, they also underscore the urgency of addressing the legal questions that accompany such rapid adoption.
Copyright law grants creators exclusive rights to their original works, including the right to reproduce, distribute, and create derivative works. In the AI context, this framework is being tested in ways it was not designed to handle.
The central question is whether AI training constitutes copyright infringement. AI companies argue that training models on copyrighted material is “transformative”—the model learns statistical patterns rather than reproducing the works themselves. Rights holders argue that using protected works without permission or compensation violates their exclusive rights.
The Authors Guild and co-plaintiffs, including authors David Baldacci, John Grisham, George R.R. Martin, and Jodi Picoult, have filed for summary judgment in their class action against OpenAI and Microsoft. They argue that “OpenAI built the foundations of its business on mass piracy,” citing the reproduction of book content from pirate sites like Library Genesis.
OpenAI and Microsoft counter that the “alleged use was fair because it was highly transformative” and that ChatGPT “does not display copies” of books. They claim a regurgitation rate of 0.00007%, meaning the model rarely produces word-for-word sequences.
The case, In re OpenAI, Inc. Copyright Infringement Litigation, is before Judge Sidney Stein in the Southern District of New York. A ruling could shape not only the economic rights of writers and publishers but also the development of generative AI itself.
The training data controversy extends far beyond books. Films, television scripts, images, music, and performances are all part of the datasets used to train AI models—often without the knowledge or consent of the creators.
Rights-holder concerns center on unauthorized use, loss of control, and competition from AI-generated content. The entertainment industry has watched as AI tools reproduce voices, insert likenesses into content they never agreed to appear in, and generate fake endorsements. These are “not fringe occurrences; they are an accelerating daily reality,” according to legal analysis.
AI-company arguments rest on fair use principles. OpenAI maintains that training is transformative because models learn patterns rather than offering the training works as substitutes. The U.S. government has even urged the court to reject a rule treating model training on copyrighted text as infringement, emphasizing industrial policy, competition, and national security.
The fair use inquiry is fact-specific. A court need not adopt a categorical rule that all AI training is lawful or that all AI training is infringing. The outcome will likely depend on how courts weigh the four fair use factors—particularly the purpose of the use, the amount used, and the effect on potential markets.
Deepfakes—realistic synthetic video and audio created using AI—present a different set of legal challenges. Unlike copyright disputes over training data, deepfakes directly implicate an individual’s identity, dignity, and right to control how their face, voice, and persona are used.
For Hollywood talent, the concerns are concrete. Actors have seen AI tools reproduce their voices in fake podcasts, insert their likenesses into content they never agreed to appear in, and generate fake endorsements of products they have never promoted. Musicians have watched AI-generated songs mimic their voices and styles.
The legal framework for addressing these harms varies by state. Washington became the latest state to expand its right of publicity law to cover “forged digital likenesses,” joining about a dozen other states including California, New York, Tennessee, and Louisiana. New York has enacted legislation requiring disclosure when advertisements feature “synthetic performers”—AI-generated likenesses of human beings.
At the federal level, the NO FAKES Act would create a federal right protecting voice and visual likeness against unauthorized digital replicas. The bill, introduced as S. 4591 in May 2026, has bipartisan support and was ordered reported to the full Senate in June 2026. It is not yet law.
The entertainment industry is increasingly confronting questions about digital replicas and synthetic performers. These terms describe different concepts:
A digital replica is an AI-generated reproduction of a real person’s likeness or voice, typically created with their consent under contractual terms. A synthetic performer is an entirely AI-generated character with no human counterpart. A human actor remains a person whose performance is captured and protected under existing law.
The 2026 SAG-AFTRA TV/Theatrical agreement strengthened protections for performers. Beyond prior notice-and-bargain requirements, producers must now demonstrate that a synthetic performer brings “significant additional value” compared to a human performer or digital replica before using it in a human role.
Contractual questions remain complex: How long does consent last? What uses are permitted? Can permission be revoked? What happens after death? The NO FAKES Act would set federal default rules, including a 10-year maximum license duration for living individuals and a 5-year limit for minors, with renewal requirements for post-mortem rights.
AI voice cloning has become one of the most immediate concerns for performers. Voice actors have discovered their voices used in content they never agreed to, from commercials to audiobooks to video games.
Legal protection for voice varies by jurisdiction. Right of publicity laws in many states protect voice alongside name and likeness. The NO FAKES Act would create a federal voice right, covering both living and deceased individuals, and would apply to “highly realistic” AI-generated replicas that are “readily identifiable” as a real person.
The legislation includes exceptions for news, commentary, criticism, parody, and satire—acknowledging the tension between protecting identity and preserving free expression. It also includes a notice-and-takedown framework for online platforms, with safe harbor protections for compliant services.
AI’s impact extends across the creative workforce. Screenwriters face questions about whether their scripts can be used to train AI models. Directors confront AI-generated visual effects and virtual production. Editors, VFX artists, and animators see AI tools automating tasks that were once human labor.
The 2023 WGA and SAG-AFTRA agreements established foundational AI protections, including digital replica consent requirements and restrictions on using writers’ material for AI training. The 2026 negotiations built on that framework, with SAG-AFTRA focusing on performer likeness, voice, and synthetic performances, while the WGA addressed literary material, training practices, and disclosure obligations.
These agreements demonstrate that AI governance has become an established collective bargaining issue. But they also leave open questions about enforcement, jurisdiction, and how to address AI developments that emerge between contract cycles.
Liability in AI-generated content disputes is a complex question with no simple answer. Potential responsible parties include:
Liability can depend on the legal theory invoked—copyright, trademark, right of publicity, defamation, contract law, consumer protection, or privacy law. A recent Ninth Circuit decision in Doe v. GitHub clarified limits on DMCA liability for AI-generated code, holding that merely generating code resembling a plaintiff’s work without removing copyright management information does not itself establish a DMCA violation.
The court distinguished between claims concerning a model’s inputs (removing information from training data) and its outputs (reproducing protected material without attribution). The decision provides important guidance as courts navigate the boundaries of AI liability.
The regulatory landscape is evolving rapidly at both state and federal levels.
State laws: Washington’s amended right of publicity law, effective June 10, 2026, creates a cause of action for unauthorized use of “forged digital likenesses.” The law includes civil penalties of $3,000 plus actual damages and profits. New York’s AI disclosure law, effective June 9, 2026, requires clear labeling of advertisements featuring synthetic performers.
Federal legislation: The NO FAKES Act (S. 4591) would create a federal right protecting voice and visual likeness against unauthorized digital replicas. The bill has 15 cosponsors—8 Republicans and 7 Democrats—and was ordered reported to the full Senate in June 2026. It remains pending legislation.
Court cases: In re OpenAI, Inc. Copyright Infringement Litigation is a multidistrict litigation consolidating lawsuits from authors, newspapers, and other copyright owners against OpenAI and Microsoft. Summary judgment motions were filed in September 2026, with a ruling expected to address the central fair use question.
The outcome of copyright and AI disputes could fundamentally alter the entertainment industry’s structure. If courts rule that training on copyrighted works requires licensing, AI companies would need to negotiate with studios, publishers, and creators—potentially creating new revenue streams but also increasing costs.
If courts uphold fair use broadly, rights holders could face a landscape where their works are used without compensation to train systems that may eventually compete with them. The Authors Guild has argued that “OpenAI’s GPT models pose an existential threat to those who write and publish books.”
Possible models for the future include licensing frameworks, revenue sharing, opt-in consent systems, and contractual AI restrictions. These are not established industry rules but approaches being discussed and tested in negotiations and litigation.
Potential Rewards
Potential Risks
Both sides of the ledger are real. The industry’s challenge is navigating them simultaneously.
The legal landscape is still evolving. Future court decisions, legislation, and union agreements will determine how AI is integrated into entertainment production—and who benefits from it.
Brandt’s observation at the L.A. Law Summit captures the moment: cultural change has outpaced legal change, leaving the industry to navigate uncertainty with existing tools while fighting for new ones. “We have to lean in, but we have to keep fighting for the rights of the content creators and intellectual property,” she said.
That dual mandate—embrace the technology while protecting the people who create—will define Hollywood’s AI era.
How is AI being used in Hollywood?
AI is used in visual effects, pre-production, localization, dubbing, editing assistance, and virtual production. Some productions have used AI to generate backgrounds and environments, reducing costs and production time. AI tools are also used for concept art, script analysis, and marketing.
What are the biggest legal risks of AI in Hollywood?
The major risks include copyright infringement from training AI models on protected works, unauthorized use of celebrity likeness and voice through deepfakes, liability questions when AI-generated content causes harm, and uncertainty about ownership of AI-created material.
Can AI-generated movies be copyrighted?
Copyright treatment depends on human authorship. Under U.S. law, works lacking human authorship may not qualify for copyright protection. The degree of human creative contribution matters. AI-assisted works may be protectable if a human contributed sufficient original expression.
Can AI use an actor’s face without permission?
State right of publicity laws generally prohibit unauthorized commercial use of a person’s likeness. Washington, California, New York, Tennessee, and other states have expanded these laws to cover digital replicas. The federal NO FAKES Act, if enacted, would create a national framework.
What are AI digital replicas?
Digital replicas are AI-generated reproductions of a real person’s likeness or voice. They may be created with consent under contractual terms or without authorization. The 2026 SAG-AFTRA agreement strengthened consent and compensation requirements for digital replicas.
Can AI clone a celebrity’s voice?
Voice cloning without permission may violate right of publicity laws in many states. The NO FAKES Act would create a federal voice right covering living and deceased individuals. Legal protection varies by jurisdiction.
Who owns AI-generated content?
Ownership is legally unsettled. If human authorship is insufficient, the work may not be copyrightable. If a human contributed sufficient creative expression, copyright may apply. Ownership in the AI company, the user, or no one is possible depending on the circumstances.
Why is copyright important for Hollywood AI?
Copyright determines who controls creative works and who profits from them. If AI companies can train on copyrighted material without licensing, rights holders lose control and compensation. If licensing is required, new markets could emerge—but at potentially higher costs for AI developers.