Who Owns a Performance When AI Helps Create It?

A performance used to be relatively easy to define. An actor learned the lines, stepped in front of the camera, delivered the scene, and received payment under an agreement that established how the performance could be used. A singer recorded a track, a voice actor brought an animated character to life, or a stunt performer executed a carefully rehearsed sequence. Technology has always influenced these creative processes, from editing and visual effects to motion capture and digital compositing. Artificial intelligence, however, is changing the questions that follow a performance after it has been recorded. A studio can now use AI-assisted tools to modify an actor’s appearance, reproduce characteristics of a voice, translate dialogue, or generate new material based on a person’s existing performance. That raises a complicated legal question: if a human supplies the original performance but a machine generates some of what audiences ultimately see or hear, who owns the result? The answer depends on several distinct rights, including copyright, a person’s rights over their voice and likeness, contractual permissions, and any applicable union agreement or employment law. These rights can overlap, but they are not interchangeable. Owning a recording does not necessarily mean having unlimited permission to recreate the performer. Having permission to use someone’s likeness does not automatically establish ownership of every new element generated with AI. And using AI in a production does not mean that the human creative contributions disappear. As entertainment companies experiment with these technologies, the legal challenge is to determine which permissions are required, how far those permissions extend, and what happens when a performance is used in a way that was never clearly contemplated when the original agreement was signed.

The Difference Between a Performance and a Digital Replica

One of the first distinctions in the AI debate is the difference between recording a performance and creating a digital replica that can produce new material. A traditional recording captures something a performer actually did: a line of dialogue, a facial expression, a dance movement, or a musical passage. A digital replica can be designed to reproduce identifiable characteristics of that performer, potentially generating new dialogue, movements, or expressions that the person never performed in that form. The replica may draw on scans, recordings, motion-capture data, voice samples, or other material. The legal significance is that the technology can extend beyond preserving an existing performance. It can create a new one using the recognizable identity of a real person.

The U.S. Copyright Office examined this issue in Part 1 of its report on copyright and artificial intelligence, published on July 31, 2024. The report addressed digital replicas: digitally created or manipulated audio, video, or images that realistically depict an individual. The Office identified gaps in existing legal protections and recommended that Congress consider a federal law addressing unauthorized digital replicas. Its analysis emphasized that the issue extends beyond conventional copyright, touching on personal identity, privacy, unfair competition, consumer protection, and fraud. <Cite refs={[“turn905014search4″,”turn905014search33”]}/> The distinction matters because a person’s voice or face may be central to their professional value without being protected in exactly the same way as a copyrighted screenplay, photograph, or sound recording. The relevant rights depend on the material involved, the jurisdiction, and the circumstances of its creation and use. An actor’s performance may be captured in a copyrighted film, while the actor’s identifiable voice and likeness raise additional legal concerns. A company may have rights to distribute the film without automatically possessing every right needed to generate an entirely new performance from the actor’s digital identity. That is why the phrase “we own the footage” cannot, by itself, settle every question about AI use.

Who Owns the Copyright When AI Is Involved?

Copyright is one part of the legal picture, but it is not a universal ownership system for everything associated with a performance. Under U.S. law, copyright generally protects qualifying original expression fixed in a tangible medium. Depending on the circumstances, that expression may include a script, a recorded audiovisual work, a musical composition, a sound recording, or sufficiently original human-authored elements in a larger project. The ownership of those rights can be determined by contracts, employment relationships, work-made-for-hire rules, and other applicable legal principles.

The Copyright Office addressed the copyrightability of AI-generated outputs in Part 2 of its report, released on January 29, 2025. It concluded that existing copyright principles can apply to AI-assisted works, but protection depends on sufficient human authorship. Human-authored expression can remain protected when AI is used as an assistive tool or when generated material is incorporated into a larger human-created work. By contrast, simply supplying prompts to a system is generally insufficient on its own to establish copyright ownership of the resulting output. <Cite refs={[“turn905014search8″,”turn905014search0”]}/>

Consider a film in which a performer records a scene, after which AI software alters the lighting, modifies the background, or helps generate a visual effect. The use of AI does not automatically erase the copyright in the underlying film or the human-authored elements. The more difficult questions arise when the technology generates expressive material that the performer never actually delivered, such as new dialogue in a cloned voice or a substantially altered facial performance. Who owns the new material? Which rights cover the use of the original performance as input? Does the contract permit the new version to be created and distributed? Those questions cannot be answered simply by identifying who operated the software. They require an examination of the creative contributions, the relevant agreements, and the rights attached to each element. Copyright in a resulting work and permission to use a particular person’s identity are separate issues. A production might have a copyright interest in an audiovisual work while still needing additional authorization to use a performer’s digital replica in a particular way.

Consent Is More Than a Signature on a Contract

Consent has become one of the central issues in entertainment’s AI negotiations because permission to perform in one production does not necessarily mean permission to generate unlimited future performances. An actor might agree to be scanned for a specific scene, for example, without expecting the scan to become a reusable digital asset for unrelated projects. A voice actor might record dialogue for one animated character but not agree to have their voice generate new lines indefinitely. A performer may also have different expectations about whether their likeness will be used for ordinary editing, a foreign-language version, a sequel, promotional content, or a scene that changes the meaning of the original performance.

The U.S. Copyright Office’s digital-replicas report highlighted the potential harm of unauthorized realistic depictions and recommended stronger protection. Meanwhile, entertainment unions have pursued contract provisions designed to clarify consent, disclosure, and compensation. SAG-AFTRA’s published resources explain that digital-replica use is addressed through specific contractual provisions and that performers should understand the difference between replicas created with their participation for a particular project and replicas created independently of a particular project. <Cite refs={[“turn905014search4″,”turn905014search6”]}/>

The practical issue is whether a performer has enough information to make a meaningful decision. A clause that broadly authorizes a studio to use a person’s likeness may leave important questions unresolved if it does not clearly explain the intended uses, duration, compensation, or ability to create new material. A performer may be comfortable with a digital alteration that supports a scene already filmed but object to having a replica generate new dialogue or appear in a separate production. The legal and contractual framework must account for those differences rather than treating every digital use as identical.

Consent can also be complicated when a performer has died. A production may wish to include a deceased actor through archive footage, visual effects, or an AI-assisted digital replica. The use of existing footage and the generation of a new performance are not necessarily the same legal activity. Depending on the jurisdiction and applicable contracts, authorization may involve an estate or other authorized representative. In a statement issued on March 20, 2026, SAG-AFTRA addressed reports about the planned digital replication of Val Kilmer for the film As Deep as the Grave. The union said the applicable collective bargaining agreement and state law required consent from the estate and stated that it understood the family’s consent had been obtained. It also emphasized that digital-replica use must be transparent, properly authorized, and consistent with contractual and legal requirements. <Cite refs={[“turn905014search3”]}/>

The broader principle is that a recognizable identity should not be treated as an unlimited creative resource simply because recordings or scans exist. The exact legal requirements vary, but consent needs to be evaluated in relation to the specific use being proposed.

What Entertainment Contracts Need to Clarify

Contracts are increasingly important because technology can develop faster than standard employment language. An agreement negotiated for a traditional performance may not anticipate a future system capable of generating dozens of new performances from the same source material. Without clear terms, performers and producers may disagree over whether a permission covers a particular use or whether a new agreement is necessary.

Useful contractual terms can address the purpose of a scan, the type of digital replica being created, the permitted projects, the duration of use, compensation, storage and security, and whether further consent is required when the intended use changes. Contracts may also need to explain how replicas can be transferred if a production company is sold or if rights move to another distributor. If a performer leaves a project, a contract should make clear what happens to any digital assets created during the engagement.

SAG-AFTRA’s published contract materials demonstrate how detailed these questions can become. Its 2025 commercials contract guidance, for example, describes requirements for consent before creating or using certain digital replicas, reasonably specific descriptions of intended use, and compensation when a replica generates a performance. The guidance also addresses security, retention, and deletion of digital-replica assets under the applicable agreement. <Cite refs={[“turn905014search13″,”turn905014search31”]}/>

These provisions matter because consent and compensation are not the same question. A performer might authorize a use but still be entitled to additional payment under the applicable agreement. Conversely, a payment does not necessarily cure the absence of required permission. A clearly drafted contract can establish what the parties agreed to, reduce uncertainty, and provide a process for negotiating uses that were not covered by the original terms. It cannot eliminate every legal dispute, but it can make the boundaries of the relationship more explicit.

The 2026 Contract Debate: Digital Replicas and Synthetic Performers

The legal conversation is also moving beyond recognizable replicas of existing actors. Some AI-generated characters are designed to resemble a specific individual, while others are synthetic performers who do not represent a particular identifiable person. The distinction matters because the legal concerns may differ. A digital replica raises questions about the use of a real person’s identity and performance. A synthetic character may raise separate questions about copyright, the use of training material, labor displacement, and whether the production is replacing work traditionally performed by humans.

SAG-AFTRA’s published information about its 2026 TV/Theatrical Agreement describes protections addressing digital replicas and synthetic performers. The union’s summary includes provisions concerning consent, digital scans, biometric data, digital alterations, foreign-language dubbing, and security of replicas. It also describes commitments related to synthetic performers, including a stated preference for human performances and conditions intended to govern when synthetic performers may be used under the agreement. <Cite refs={[“turn905014search9”]}/>

These provisions should be understood in their contractual context. A collective bargaining agreement applies to covered work and parties; it does not automatically establish identical rules for every independent production, country, or category of creative labor. Nor does a union contract resolve every question about copyright ownership or AI training. But such agreements can provide concrete rules where broad legal principles leave room for uncertainty. They can also create a practical standard for notice, bargaining, consent, and compensation within the productions they cover.

The distinction between digital replicas and synthetic performers illustrates why the legal debate is unlikely to be settled by a single rule. Different technologies create different risks, and different uses may require different contractual safeguards. A background adjustment, an AI-assisted voice translation, a new line delivered by a digital replica, and a fully synthetic character are not necessarily equivalent situations. The more precisely a contract distinguishes among them, the easier it becomes for the parties to understand their rights and obligations.

What Happens to Creative Credit?

Ownership and consent are only part of the issue. Creative credit can become more complicated when a production uses AI to generate or alter parts of a performance. If an actor’s face is digitally modified, a voice is synthesized, or an AI system generates new movement based on recorded material, audiences may have difficulty understanding who contributed what. The actor may have supplied the original expression, the director may have shaped the performance, the visual-effects team may have built the digital transformation, and technical specialists may have used AI tools to generate additional material.

There is no single credit rule that automatically answers every question about these contributions. The applicable contract, industry practice, and nature of the work may determine how credits are assigned. In some cases, the contribution may be primarily technical. In others, it may substantially affect a character’s performance or the meaning of a scene. A production can also involve multiple layers of creative work that do not fit neatly into one category.

Clear credit practices can help audiences and collaborators understand how a project was made. They can also reduce the risk that human contributions are overlooked simply because a digital tool played a visible role. For performers, credit may affect professional reputation and future opportunities. For visual-effects artists, animators, editors, and sound designers, recognition can be important to the value of their work. For studios, transparent production records can help document which elements were generated, modified, or supplied by different contributors.

The challenge is not to assign every technical operation the same status as a starring performance. It is to establish fair and understandable standards that reflect the actual contributions involved. As AI becomes more deeply integrated into production, the industry may need more detailed ways to record creative decisions and distinguish between editing an existing performance and generating new expressive material.

The Rights of Voice Actors and Musicians

Voice performers and musicians face particular questions because their voices can be reproduced and modified digitally. A voice is both a creative instrument and a recognizable aspect of personal identity. AI systems may be used to create alternate-language dialogue, adjust vocal delivery, generate additional lines, or produce audio that resembles an existing performer. Some uses can help a production reach new audiences or improve accessibility. Others may raise concerns about unauthorized imitation, compensation, or the possibility of generating material that the performer would never have agreed to record.

SAG-AFTRA’s materials on sound recordings describe collectively bargained protections requiring consent before a recording using digital replication of an artist’s voice is released under the applicable agreement. The union states that consent must be in a separate writing containing clear and conspicuous language and reasonably specific information about the intended use. <Cite refs={[“turn905014search12”]}/>

These safeguards illustrate the importance of separating the permission to record a voice from the permission to use AI to generate new performances. A recording contract may address a particular song, session, or commercial release without necessarily answering every question about future synthetic uses. The same principle applies to dubbing and translation. An AI-assisted translation might preserve the original performer’s voice characteristics, but the agreement still needs to address the scope of that use and any applicable compensation.

For performers, a practical contract review should identify whether a replica can generate new content, whether the permitted use is limited to a specific project, whether additional consent is needed for later uses, and how payment is calculated. For producers, clear authorization helps establish a defensible process and reduces the likelihood that an otherwise useful technology becomes the center of a dispute. The legal details vary by contract and jurisdiction, so these questions should be evaluated with qualified legal advice rather than assumptions about what a general permission clause means.

AI Training Raises Another Layer of Intellectual Property Questions

The ownership debate becomes even more complex when AI systems are trained using existing creative works. A digital replica may use material from a particular performer, but a generative model may also be trained on large collections of text, images, audio, or video. These are related but different legal issues. Permission to use an actor’s likeness in a particular production does not automatically resolve whether a separate model may be trained on copyrighted recordings or other protected material.

The U.S. Copyright Office’s ongoing AI initiative has examined several distinct questions, including the copyrightability of AI-generated outputs, digital replicas, and the use of copyrighted works in AI training. Its reports emphasize that these issues require separate legal analysis rather than one universal answer. <Cite refs={[“turn905014search0″,”turn905014search1”]}/>

For entertainment companies, this distinction has operational consequences. A studio may possess a large library of films and recordings, but that does not necessarily mean every asset can be used for every AI purpose. Contracts with performers, writers, composers, production vendors, and distributors may contain terms that affect whether particular materials can be used for training or development. Some agreements may require additional consultation or consent; others may impose restrictions on copying, licensing, or commercial use. The specific rights depend on the relevant contracts and law.

Companies that plan to develop or license AI systems therefore need to track the provenance of their training material and the permissions associated with it. Creators need to understand what they are authorizing when they supply work to a platform or sign an agreement that includes AI-related language. Better documentation cannot settle every legal dispute, but it can help identify where permissions exist, where restrictions apply, and where further negotiation is needed.

What Producers and Performers Should Ask Before Signing

As AI-assisted production becomes more common, both sides of the employment relationship benefit from asking specific questions before work begins. A performer should understand whether a digital scan or voice sample is being collected, what it will be used for, how long it will be retained, and whether it can generate new performances. The agreement should identify whether the intended use is limited to a particular film, episode, commercial, game, or other project, and whether additional uses require fresh authorization.

Compensation also deserves careful attention. If an AI-generated performance replaces or supplements work that would otherwise require a performer to return to the studio, the parties should clarify what payment applies under the contract. They should determine whether the replica can be used in sequels, promotional material, localized versions, or unrelated productions. They should also consider how the agreement handles security, access, transfer of ownership, and deletion of the underlying digital asset.

Producers, meanwhile, should establish a process for confirming that the necessary permissions have been obtained before a replica is used. That process may involve legal teams, casting departments, production managers, effects vendors, and distributors. Documentation should be clear enough that the people operating the technology know what they are authorized to do. A production should not rely on the assumption that an asset created for one purpose can be reused for any other purpose simply because the technology makes reuse possible.

These considerations are general information, not a substitute for legal advice. Applicable rules differ across jurisdictions, employment categories, union agreements, and contractual arrangements. But asking these questions early can help prevent misunderstandings and make creative collaboration more sustainable.

The Legal Future of Performance in an AI-Assisted Industry

AI does not make traditional performances disappear, but it changes the relationship between the person who performs, the material that is recorded, and the company that distributes the result. A single scan or recording can potentially support uses that extend far beyond the original session. That capability makes consent, compensation, ownership, credit, and data security more important—not less.

The developing legal framework reflects several separate concerns. Copyright law determines which creative expression is protected and who may control it. Rights relating to voice and likeness address the use of a person’s identity. Contracts define the permissions exchanged between performers and producers. Union agreements can establish specific safeguards for covered work. Privacy, consumer-protection, and other laws may also apply depending on the circumstances. No single category answers every question.

The central challenge is to preserve the ability to innovate without treating human performance as an unlimited digital resource. AI can help create new effects, expand access to international audiences, and support production workflows. But those benefits do not eliminate the need to respect the people whose work, voices, faces, and movements make entertainment valuable in the first place.

The question of who owns an AI-assisted performance is therefore not simply a debate about whether machines can create. It is a question about how creative rights should be allocated when the production process becomes more complicated. The answer will depend on the material involved, the permissions granted, the applicable law, and the agreements negotiated by the people doing the work. As the technology advances, the clearest path forward is not to assume that every new use is already covered by an old contract. It is to define those uses carefully, obtain meaningful consent where required, compensate people according to the applicable agreements, and make the contributions behind a performance easier to understand. That approach gives entertainment companies room to experiment while helping ensure that the people behind the performances are not left out of the future they helped create.

The Judge