Digital IdentityLegal Analysis

Who May Speak for the Dead?

Consent, Digital Inheritance, and the Governance of Posthumous Artificial Intelligence

Published October 14, 2026

Last revised October 14, 2026

Brass microphone, archival papers, and a golden waveform against a dark background — Conceptual illustration

Legal analysis: This article examines legal principles and proposed approaches. It is not advice about a particular situation.

Author’s Note: The author is an attorney admitted in Florida and Alabama and the founder of PersonaVault™. PersonaVault™, Digital Likeness Trust™, and Digital Afterlife Agent™ are names used for the planning framework discussed here. The author has a commercial interest in developing services based on that framework. The legal analysis does not depend on the use of a branded service.

Abstract

Posthumous artificial intelligence presents an administrative problem that ordinary digital-asset transfers do not resolve. Access to recordings, ownership of intellectual property, an individual’s lifetime consent, and a fiduciary’s authority to approve future projects are distinct. This Article proposes a method for coordinating them: preserve evidence of permission, define categories of authorized creation, assign accountable decision-making powers, and translate restrictions into enforceable vendor obligations. Examples involving Anthony Bourdain, James Earl Jones, estate-authorized voice licensing, and George Carlin demonstrate why those distinctions matter. Using Florida trust and digital-access law as a principal illustration, the Article examines who may act when beneficiaries prefer revenue to the creator’s restrictions. Its contribution is a governance approach for authorized projects, rather than a new property right in personality. It also distinguishes private obligations from claims against independent speakers, whose expressive freedom an estate plan cannot simply override.

I. Authentic Words, a Performance That Never Occurred

In the documentary Roadrunner, an artificial recreation of Anthony Bourdain’s voice delivered words he had written. The words were his; the recorded performance was not. Disclosure of the technique prompted controversy about how the documentary represented its subject.¹

The episode exposes a distinction that becomes increasingly consequential as generative technology improves. Authentic source material does not make every resulting representation authentic. A recording can sound like a historical document while embodying a creative decision made after the speaker’s death.

Estate planning must now confront that distinction. Traditional planning asks who receives what a person leaves behind. Posthumous AI introduces a further question: who may authorize what will subsequently be created in that person’s name?

A conversational replica can answer questions the deceased never considered. A synthetic performance can place the individual in a production the individual never reviewed. A voice model can read unfamiliar text with the apparent intimacy and authority of a recognizable human voice.

These possibilities are not inherently contrary to the individual’s wishes. Someone may welcome a continuing creative career, an educational resource, or an interactive family archive. Another person may permit preservation while prohibiting all newly generated expression. The legal problem is how to distinguish those choices and administer them after the person can no longer clarify their meaning.

This Article argues that planning should treat permission for future creation as an administrative responsibility separate from the transfer of source materials. A workable arrangement should preserve the evidence on which an approval rests, define who may approve or reject a use, and provide a legally effective means of reviewing that decision.

II. The Contribution: Governing Authorized Uses

Scholarship on digital inheritance already challenges assumptions about what passes at death. David Horton’s treatment of “indescendibility” examines interests that cannot be transmitted through ordinary succession, including questions involving digital accounts. Its relevance here is foundational: an instruction to transfer an interest does not necessarily establish that the interest is transferable.²

Anita Allen and Jennifer Rothman’s account of postmortem privacy supplies a complementary caution. Their analysis considers interests that a property-centered approach can obscure, while recognizing the living’s legitimate engagement with the dead. Treating identity exclusively as inheritable property risks overlooking both personal interests and expressive freedom.³

The present proposal addresses a narrower administrative problem: how should people organize projects they wish to authorize, or restrictions they wish participating parties to observe?

The proposal is administrative in scope. It does not attempt a comprehensive survey of digital-replica legislation or advocate unlimited postmortem control. Its question is what a person and participating institutions can arrange now so that an authorized project remains faithful to defined purposes after death.

This inquiry differs from asking whether the law should create broader rights against unauthorized replicas. The Copyright Office recommended federal legislation addressing gaps in existing protections in its 2024 digital-replicas report. Whatever the appropriate public-law response, authorized projects still require a process for interpreting permission, clearing relevant rights, selecting vendors, and supervising performance.⁴

The proposed contribution is an approval process linking four questions that otherwise tend to be handled separately. Before a project proceeds, its administrator should document access, relevant ownership and licenses, the scope of lifetime instructions or later authorization, and the source of the administrator’s power. That record should then guide the vendor agreement and remain available for review. This is a proposed planning method within existing legal structures, not a claim that current law universally requires such a process.

III. Four Questions Behind a Digital Replica

Consider a novelist who leaves recordings, correspondence, and unpublished manuscripts. One child proposes a searchable archive. Another wants a conversational simulation. A publisher offers to finance the simulation if it may also generate new stories under the novelist’s name.

The proposals involve overlapping materials but different authority.

Access concerns who may obtain and retain the materials. Ownership concerns copyright and other relevant interests. Consent concerns the uses the individual approved and any additional permissions required. Fiduciary authority concerns who may interpret instructions, approve transactions, and supervise implementation.

Possession of the files does not conclusively answer any of these questions.

Federal copyright law expressly distinguishes ownership of a material object from ownership of copyright. Copyright interests may pass through succession or transfer, but transfers of copyright ownership generally require a signed writing unless they occur by operation of law. Inheriting a photograph therefore does not itself transfer the photographer’s copyright.⁵

Similarly, an individual’s permission to recreate a voice does not necessarily resolve rights in a recording used for that purpose. Conversely, rights in a recording do not necessarily establish authority to exploit the depicted individual’s identity. Each proposed use requires examination of the relevant interests, licenses, and applicable limitations or exceptions.

Florida’s fiduciary-access statute addresses another part of the problem. Chapter 740 provides disclosure procedures, distinguishes communication contents from other digital assets, and recognizes specified directions made through online tools. It also preserves relevant limitations arising from terms of service and other law.⁶

The planning implication is that permission to retrieve correspondence should be coordinated with instructions concerning its subsequent use. Access needed to administer an estate does not expressly confer a general power to construct an AI persona.

The novelist’s fiduciary might obtain her email while lacking permission to use it in an interactive system. Incoming correspondence may also implicate rights and interests of living people. The deceased’s directions cannot be assumed to dispose of those interests.

IV. Advance Consent and Authorization by Survivors

James Earl Jones provides an affirmative example of planning before death. Reporting in 2022 described his approval of the use of archival recordings to recreate Darth Vader’s voice through AI. The documented approval demonstrates that a performer can participate in decisions about synthetic continuation while alive.⁷

The example supports a limited proposition. Public reporting does not establish every contractual term, and it does not show that Jones employed a trust. It does illustrate why the scope of permission matters. Authorization associated with a particular character should not be assumed to authorize unrelated endorsements, personal statements, or unrestricted simulation of the performer.

Estate authorization presents a different question. In 2024, ElevenLabs announced partnerships with the estates of Judy Garland, James Dean, Burt Reynolds, and Laurence Olivier to make their voices available in its Reader app. The announcement demonstrates a market for estate-authorized synthetic performances. It does not itself establish what each individual approved during life.⁸

The distinction is central to this Article: authorization exercised by survivors and consent expressed during life are separate sources of permission. An estate-authorized use may be lawful even without evidence that the individual anticipated or approved it. The proposal does not make lifetime consent a universal prerequisite. It instead requires the administrator to identify which source of permission supports the project and to avoid describing a survivor’s choice as the deceased’s own.

A sound plan should preserve evidence of both. Where lifetime instructions exist, the decision-maker should be able to identify their scope. Where they do not, the arrangement should identify the source and limits of the survivor’s authority without presenting a later decision as the deceased’s own choice.

V. Consent Should Identify Categories of Creation

“Preserve my legacy” expresses an aspiration. It does not determine whether the novelist authorized synthetic interviews, new fiction, or commercial endorsements.

This Article proposes that planning instruments require express authorization for materially new categories of representation. Preservation language alone should not suffice, within the proposed framework, to authorize an open-ended system that speaks as the deceased. This is a drafting recommendation rather than a claim that current law universally imposes that rule.

Permission should distinguish replaying an authentic recording, synthesizing a performance from approved text, and generating new expression. It should also identify the audience, purpose, duration, and discretion granted to the decision-maker.

For the novelist, permission to make recordings searchable would not, under the proposed drafting approach, imply permission to publish newly generated stories under her name. The administrator would identify the proposed category, check the operative instructions, clear the necessary rights, and explain the decision in writing. If permission were uncertain, the specified review process would apply before the vendor began production.

A performer could permit future entertainment productions while prohibiting political advertising. A parent could authorize private family interaction while withholding permission for public distribution. An author could permit readings of existing works but prohibit new works presented under the author’s name.

These choices need not predict every future script or technology. They should establish intelligible boundaries and identify who may resolve uncertainty.

The Bourdain example helps explain why authenticity must also be divided into components. A system may reproduce authentic words through a synthetic voice. Another may generate new words using authentic recordings. The source, performance, and resulting claim of authorship should be documented separately.

VI. PersonaVault™ as Evidence of Permission

PersonaVault is proposed as a curated archive of source materials and associated directions. The legal analysis does not depend on a particular branded service.

The archive’s value would lie in preserving information needed to make defensible decisions: provenance, ownership, applicable permissions, restrictions, and the controlling version of the individual’s instructions.

It should distinguish operative directions from explanatory material. A recorded conversation about preferences may assist interpretation without satisfying the requirements for amending a governing instrument. Updates should therefore follow a defined process coordinated with the relevant legal documents.

The archive should also identify uncertainty. A recording with unclear ownership should not silently become an approved production asset. Material containing third-party information may require review before disclosure or reuse.

Preserving these distinctions would help a fiduciary explain why a proposed use was approved or rejected. PersonaVault would supply evidence for administration; it would not certify that every desired use is lawful.

VII. The Case for a Digital Likeness Trust™

A trust is most useful when decisions will recur: multiple licensing proposals, changing technology providers, continuing expenses, and beneficiaries whose interests may diverge.

Simpler objectives may require simpler arrangements. A family seeking only to retain recordings may need access instructions and funded preservation. A single production may be governed chiefly through a license. An operating entity may suit a continuing commercial enterprise.

A will can establish a testamentary trust, and an existing revocable trust can contain specialized provisions. A separately titled Digital Likeness Trust is therefore an organizational option, not a legal prerequisite.

Florida’s trust-creation requirements remain applicable, including capacity, intent, trustee duties, and a definite beneficiary or an applicable exception. The arrangement must identify property or enforceable interests that can actually be administered.⁹

Relevant interests might include owned copyrights, contractual rights, archive assets, and properly established licensing authority. Their scope and transferability require separate analysis.

The trust’s contribution is continuing administration under defined purposes. Its name creates no additional right in personality. Without funding and effective instruments, it may offer little beyond an expression of preference.

The settlor should retain a defined route for revising instructions while able to do so. At death, the administrator should identify the operative versions of the archive, trust provisions, and licenses, rather than infer a final instruction from an undated recording. Funding should account for the practical costs of storage, legal review, and administering the restrictions.

VIII. When Beneficiaries Prefer Revenue

Suppose the novelist expressly prohibited advertising uses, but an endorsement would generate substantial income. Every beneficiary supports the transaction.

The conflict should not first be discovered when the offer arrives. The governing instrument should specify whether income production is subordinate to particular restrictions, what preservation expenditures are authorized, and how administrative costs affect continuation of the arrangement.

Florida’s duties of loyalty and impartiality make beneficiary interests legally significant. Those duties cannot be replaced with a vague instruction to protect the deceased’s reputation.¹⁰

Enforcement incentives matter because a restriction may become least secure when it is most costly to honor. Independent oversight can help only if the overseer has effective powers, information, and a legally recognized route to relief. The instrument cannot confer judicial standing merely by declaring that someone represents the deceased’s wishes.

The instrument should identify material purposes and procedures for resolving disputes. It should also contemplate changed circumstances, including technological obsolescence and uneconomic administration.

A directed trust offers one possible division of responsibility. The instrument could assign an independent trust director authority to approve or reject specified categories of replica use while leaving investment and distribution functions with the trustee. Florida permits powers of direction defined by the trust’s terms and imposes duties and liabilities through its directed-trust provisions. The role should specify the decisions reserved to the director, the information required, successor selection, conflicts, and removal. A label such as “trust protector” is insufficient without operative powers.¹¹

Judicial enforcement remains a separate inquiry. Florida permits intervention when an interested person invokes the court’s jurisdiction or as otherwise provided by law, and treats a trust director as a trustee for specified purposes unless the terms provide otherwise. Information rights depend on statutory rules and valid trust terms. The plan should identify who receives reports, who can exercise a defined power, and who may seek judicial relief under the applicable law; it should not assume that every adviser can sue.¹²

A workable arrangement should require a written record for sensitive approvals, give the designated reviewer timely access to that record, and fund independent advice and any legally available enforcement. Compensation should not make that reviewer dependent on approving more licenses. These measures do not guarantee litigation or a particular outcome. They reduce the chance that unanimous financial incentives will leave a restriction unexamined.

Permanence cannot be guaranteed. Florida permits judicial modification in specified circumstances and preserves judicial authority over modification and termination. The framework must operate within those limits.¹³

Its defensible objective is accountable administration of clearly expressed purposes, subject to law.

The novelist example also reveals a limit. If the purported restriction is only an informal preference, or if the governing law permits modification on the facts presented, independent oversight cannot make it immutable. The proposal therefore emphasizes properly expressed purposes, valid powers, and reviewable decisions rather than a promise of permanent obedience.

IX. Professional Representation and Contractual Control

The estate partnerships announced by ElevenLabs illustrate why administration may require specialized expertise. Evaluating a synthetic-voice license involves permitted uses, distribution, compensation, technical restrictions, and the consequences of termination.

A Digital Afterlife Agent™ is proposed as a professional role assisting with those tasks. The title would confer no independent legal authority. Responsibilities would arise from the engagement and any valid delegation.

Florida permits prudent delegation while requiring care in agent selection, definition of responsibilities, and periodic review. That structure offers a foundation for assigning operational work while maintaining appropriate oversight.¹⁴

The engagement should identify reserved approvals, conflicts procedures, compensation, and removal rights. An agent compensated solely through licensing revenue may have inadequate incentives to recommend preservation or silence.

Vendor contracts must translate the plan into operational obligations. They should address approved inputs, permitted generation, reuse in other systems, subcontracting, security, audit rights, and service discontinuation.

Termination requires particular care. Ending future generation, deleting source files, addressing retained models, and withdrawing previously distributed works are different tasks. The contract should state which obligations apply and what verification is feasible.

Disclosure should also describe what the project actually does. A recording of authentic words performed synthetically differs from newly generated expression in a familiar voice. Labels and contractual disclosure requirements should make that distinction intelligible to the intended audience. A disclosure cannot supply missing rights or cure an unauthorized use.

X. George Carlin and the Burden of Enforcement

In 2024, George Carlin’s estate and other rights holders challenged a comedy special presented as an AI recreation of the deceased comedian. A June 18 stipulated consent judgment barred the defendants from reposting the special or using Carlin’s image, voice, or likeness on their controlled platforms without the plaintiffs’ written approval. The court also declined the parties’ request to bind unrelated third parties because no basis for doing so had been provided. The resolution is a consent judgment, not a contested merits ruling establishing a general law of digital replicas.¹⁵

For planning purposes, the episode illustrates the need for someone with authority, resources, and evidence to respond after a contested release.

A trust cannot prevent every unauthorized production. It can organize the interests supporting a claim, maintain relevant records, and allocate responsibility for obtaining advice and pursuing relief. The archive can assist in distinguishing authorized material from challenged uses.

Enforcement should remain tied to identifiable legal rights. A resemblance that offends beneficiaries does not automatically establish liability, and litigation objectives should reflect both the available claims and the trust’s resources.

XI. The Limit: An Estate Does Not Own Public Memory

Florida’s section 540.08 protects against specified commercial uses of name or likeness, recognizes written licensing authority, and provides a postmortem family fallback. It also includes exceptions and a forty-year limit.¹⁶

In Tyne v. Time Warner Entertainment Co., the Florida Supreme Court held that the statute’s commercial-purpose language did not encompass publications, including motion pictures, that did not directly promote a product or service. The court also identified constitutional concerns with broader coverage. Although Tyne did not address generative AI, it demonstrates that selling an expressive work does not alone establish liability under the statute.¹⁷

New York illustrates a different statutory approach. Its current section 50-f defines deceased personalities and performers with domicile and commercial-value requirements, expressly addresses digital replicas, and permits the transfer of recognized rights through trusts and other instruments. It also contains expressive-use exceptions, a successor-registration requirement affecting claims, and a forty-year duration. Those provisions show both the potential for succession planning and the need to examine the particular right rather than assume that every identity interest follows the same rules.¹⁸

The framework distinguishes contractual control over participating parties from claims against independent speakers. A vendor may agree to restrictions that do not bind a critic, biographer, or filmmaker.

This difference is consequential for enforcement. The administrator may have a contractual claim against a vendor that exceeds its license even where a similar independent expressive work would fall outside a publicity statute. Conversely, an outsider’s conduct requires an independent legal basis for relief. The planning record supports that analysis; it does not replace it.

This boundary is essential. Estate planning should organize authorized representations without converting the deceased’s preferences into a universal prohibition on interpretation, criticism, or historical inquiry.

XII. Conclusion

The examples reveal different questions that are too easily grouped together. Bourdain’s synthetic narration raises the distinction between authentic words and authentic performance. Jones’s approval illustrates the value of decisions made during life. Estate voice partnerships demonstrate a market for authorized continuation. The Carlin dispute illustrates the burden of responding to contested uses after release.

No single asset-transfer provision resolves all of these issues.

Access makes materials available. Ownership identifies relevant entitlements. Consent establishes permission and its limits. Fiduciary authority organizes decisions over time.

PersonaVault, a properly constituted Digital Likeness Trust, and carefully negotiated contracts could coordinate those functions, with a Digital Afterlife Agent providing specialized assistance where appropriate. Their purpose would be to make individual choices administrable: to preserve, to create, to commercialize within boundaries, or to remain silent.

The central promise of posthumous AI planning should be an accountable decision process supported by evidence. A future administrator should be able to explain what was inherited, what was approved, who had power to decide, and how that decision can be reviewed. Possession alone cannot supply those answers.

Sources

← All Articles & Commentary