Digital Replicas in Expressive Works: NILV Rights vs. Free Speech

The End of a Boundary That Seemed Immovable

For decades, the right to one's own image operated on a distinction that legal scholars considered structural: the commercial and the expressive were mutually exclusive categories. Using someone's image to sell a product without consent constituted an actionable wrong. Using it to tell a story — in a film, a song, a novel — was, by definition, a legitimate exercise of freedom of expression. That boundary, which Anglo-American law called the right of publicity and which European law articulated through the image rights of Article 8 ECHR and Article 7 of the Charter of Fundamental Rights of the European Union, has just been substantially challenged.

Benjamin Sheffner's article published in the Columbia Journal of Law & the Arts in 2026 documents with surgical precision the legislative process that has led the United States to cross that boundary. But the phenomenon is not American: it is global. And its legal translation in the European context — where the Artificial Intelligence Act (Regulation EU 2024/1689, hereinafter "AI Act" or "AIA") has created its own regime governing generative systems and so-called "digital replicas" — demands doctrinal analysis that goes beyond mere description of the state of legislation.

The thesis advanced here is this: the expansion of Name, Image, Likeness & Voice ("NILV") rights into the territory of expressive works constitutes a first-order normative collision, whose resolution cannot be achieved through simple legislative exemptions of blurred contours, but requires a doctrinal reconstruction of the very foundation of these rights in light of the proportionality principle and the constitutional architecture of freedom of expression.

Technology, Moral Panic, and Reactive Legislation

The historical starting point Sheffner identifies is precise: the proliferation of deepfake pornography in 2017 — with the faces of actresses and singers superimposed on other bodies in explicit scenes — generated what sociologists would call a technological moral panic. Two distinct vectors compounded that initial stimulus: the replication of actors' voices and appearances in new works without their actual participation (the viral "deepfake Tom Cruise" on TikTok; songs generated with synthetic versions of real artists' voices), and the deployment of digital replicas of politicians to attribute statements they never made.

What is doctrinally significant is not that these phenomena cause harm — they do, to varying degrees and for distinct reasons — but that legislators responded to all three with a single homogeneous instrument: expanding the right of publicity to cover expressive works. This legislative homogenization of heterogeneous situations is, precisely, one of the structural defects of the regulatory movement being analyzed.

Non-consensual intimate deepfakes are fundamentally a matter of privacy and personal dignity: the appropriate legislation concerns non-consensual intimate imagery, as addressed in the U.S. by the TAKE IT DOWN Act of May 2025. Electoral manipulation via digital replicas threatens democratic integrity and finds its response in election-specific legislation. Only the third scenario — replicating actors and musicians in new works without their consent — genuinely poses the tension between identity rights and creative freedom. Yet all three have converged in the same legislative wave, contaminating the doctrinal debate and preventing the normative instrument from being calibrated precisely to each distinct legal interest at stake.

The NILV Normative Architecture: Expansion Without Defined Constitutional Content

Sheffner's analysis reveals a legislative pattern that, viewed from European legal dogmatics, is disturbing: states including California, Tennessee, Illinois, and Montana have expanded their right-of-publicity statutes into expressive works by formally including exemptions for uses protected by the First Amendment — satire, parody, criticism, documentaries, biopics — but configuring those exemptions so indeterminately that invoking them requires, in practice, a full constitutional analysis on a case-by-case basis.

Tennessee's ELVIS Act carries this problem to its purest expression: its exemptions activate "[t]o the extent such use is protected by the First Amendment to the United States Constitution" — which is legally equivalent to saying nothing operational at all. An exemption that refers entirely to the applicable constitutional standard does not fulfill the function the legislature assigns to it: providing ex ante certainty about whether proposed conduct is lawful. A creator intending to use a digital replica of a deceased musician in a work of fiction has no reasonable way of knowing in advance whether that action is legal or exposes them to litigation.

This problem has a technical name in U.S. constitutional doctrine: chilling effect. And in the dogmatics of the ECHR and the CJEU, its functional equivalent is the requirement that any restriction on artistic freedom of expression be "foreseeable" in its consequences — a sine qua non condition for surviving the proportionality test under Article 10(2) ECHR.

The European Mirror: AI Act, GDPR, and Identity as a Complex Legal Interest

The pertinent question for the European jurist is not whether the Union's legal order has replicated the American legislative movement — it has not done so with the same intensity — but whether the current normative framework provides adequate responses to the same phenomena, and whether those responses are constitutionally more robust.

The AI Act, in force since August 2024 with phased application, contains three instruments relevant to the analysis. First, Article 5(1)(f) prohibits AI systems that generate or expand facial recognition databases through indiscriminate scraping of images, constituting a preventive restriction at the training stage. Second, Article 50(2) imposes on providers of generative AI systems — including those capable of generating voice or image replicas — the obligation to mark synthetic outputs in a technically detectable manner. Third, Recital 132 of the AIA explicitly acknowledges the tension between these obligations and artistic freedom, noting that the Regulation's application must respect "freedom of the arts and sciences" enshrined in Article 13 CFREU.

What the AI Act does not do — and here lies the structural difference from the American legislative movement — is create a sui generis right over voice or image in expressive works. The Regulation operates on systems, not on content; it regulates providers and deployers, not authors and creatives. This distinction is, from the perspective of freedom of expression, of enormous significance: a norm imposing transparency about the synthetic nature of content marginally affects the expressive act, while a norm granting NILV rights holders veto power over a work affects its very existence.

It should be added that the GDPR provides a data protection regime covering facial images and voice — special categories under Article 9 when their processing enables identification — but with explicit exceptions for journalism, artistic expression, and academic research under Article 85. This normative design — protecting without prohibiting — is the model European doctrine should champion against NILV legislative temptations.

The Constitutional Antinomy: Identity Against Creative Freedom

The doctrinal core of the problem can be stated precisely: what is the constitutional title justifying a veto power over expressive works in favor of NILV rights holders, and what weight does that title carry against the fundamental right to artistic freedom?

In U.S. law, as Sheffner notes, NILV statutes are content-based restrictions subject to strict First Amendment scrutiny: to survive constitutional review they must serve a compelling governmental interest and be designed with the least restrictive means necessary. The author identifies a plausible justification — protecting the ability of living actors and musicians to earn a living without digital replicas competing with their work — but notes that this justification does not exist for deceased individuals, whose interest in receiving remuneration for their work ceased with death.

In European law, the test is functionally equivalent though terminologically distinct: Article 52(1) CFREU requires that any limitation on fundamental rights — including the artistic freedom of Article 13 — respect the "essential content" of those rights and be proportionate. The ECtHR, for its part, has built solid jurisprudence on artistic freedom that recognizes considerable margin of appreciation for creators, particularly when a work contributes to public debate or has cultural value.

What the NILV expansion into expressive works violates, in both legal orders, is the requirement of proportionality in the strict sense: the balance between the legal interest protected — the individual's ability to control their image in creative contexts — and the cost of that protection on artistic freedom of expression is, in the majority of cases covered by new legislation, unfavorable. This is because legislators have opted for the most intensive instrument — the rights holder's veto — when less restrictive instruments were equally apt to protect the interests at stake: an obligation of transparency about the synthetic nature of the content; recognition of a mandatory equitable remuneration without veto rights; or a reversal of the burden of proof in cases of actual demonstrable harm to reputation or earnings.

Posthumous Rights as an Autonomous Problem

One of the strongest arguments in Sheffner's article, meriting independent doctrinal development, is the constitutional contestability of posthumous NILV rights over expressive works. The economic justification — protecting the ability to earn a living — disappears with the rights holder's death. The alternative justifications that legislators invoke — dignity, reputation, protection of heirs — do not withstand dogmatic scrutiny.

Dignity is a personal and non-transferable legal interest. The Restatement (Second) of Torts is categorical: "there can be no defamation of the dead." The same logic applies to the interest in controlling one's image in expressive works: that interest exists while there exists a subject capable of experiencing the harm. Heirs have their own interests — patrimonial, even sentimental — that law can protect, but those interests are not the same interests that justified protection during the rights holder's lifetime.

In European law, the morte civile of fundamental personality rights is a recognized principle. The GDPR does not apply to data of deceased persons — though Member States may regulate the matter with some latitude — and the ECtHR has consistently held that Article 8 ECHR rights are personal and extinguish with the rights holder's death. Creating a posthumous proprietary right over image in expressive works lasting decades — Tennessee's statute potentially endures indefinitely; California and the proposed NO FAKES Act suggest seventy post-mortem years — amounts to constituting a deferred private censorship, administered by heirs or corporate assignees, over cultural production involving deceased individuals.

The systemic risk this generates for film, literary, and musical production is difficult to overstate. Holders of posthumous NILV rights — often specialized corporate entities such as CMG Worldwide, which manages rights of figures including James Dean, Bette Davis, and Thomas Edison — have economic incentives to maximize the value of their asset, including by preventing or pricing-out works outside their control. This private veto power over collective cultural memory finds no plausible constitutional justification in either legal order analyzed.

Toward a Constitutionally Sustainable Digital Replica Doctrine

The thesis of this analysis is not that creators of digital replicas should operate in a normative vacuum. It is that the correct legal instrument to protect the genuinely at-stake interests is not the right of publicity expanded into expressive works, but a differentiated regime that includes at least the following elements.

First, the distinction between a competing replica and an expressive replica should be the axis of the system. A digital replica of a living actor generated to replace them in a work they could have performed — directly competing with their labor — affects a real economic interest and justifies regulatory intervention, perhaps including recognition of a mandatory equitable remuneration negotiated collectively, along the model of the sui generis database right. A replica of the same actor in a work of political satire, a historical documentary, or a fiction that takes them as a character, only marginally affects that interest and cannot justify identical treatment.

Second, transparency as the primary obligation and veto rights as a last resort. The obligation to identify synthetic content — as the AI Act already requires in Article 50(2) — protects the public without closing the creative space. Only when a digital replica causes real, accrued, and demonstrable harm not remediable through transparency should the more intensive intervention of the legal order be triggered.

Third, the extinction of rights over expressive works upon the rights holder's death, without prejudice to the protection of heirs' patrimonial interests through the general regime of intellectual property as applied to works in which the deceased actually participated. Personality rights cannot be transmitted mortis causa without mutating into something qualitatively distinct — a private censorship right — that constitutional texts do not support.

Conclusion: The Normative Revolution and Its Constitutional Limits

Sheffner concludes, with admirable restraint, that right-of-publicity law is "in the beginning stages of a revolution" whose ultimate equilibrium will take decades to reach. That is a plausible prediction. But doctrinal analysis cannot limit itself to describing the process: it must identify the constitutional limits that process cannot cross without violating the foundations of the liberal rule of law.

Those limits are clear in both legal orders. Norms granting NILV rights holders a general veto over expressive works are, in principle, disproportionate and incompatible with artistic freedom of expression. Legislative exemptions that refer entirely to constitutional analysis on a case-by-case basis fail the foreseeability requirement that both U.S. constitutional law and the European proportionality standard demand. And posthumous rights over expressive works lack sufficient constitutional foundation, both in First Amendment dogmatics and in the jurisprudence of the ECtHR and CJEU.

What technology has made possible — the hyperrealistic synthetic replication of any living or deceased individual — cannot give rise to a normative response that, in protecting legitimate but partial legal interests, sacrifices the freedom enabling democratic societies to remember, imagine, criticize, and narrate the world they inhabit. The law of digital replicas must be built on a distinction that the legislative moral panic has obscured: the difference between commercially exploiting someone's identity and telling a story in which that identity appears.

This analysis is academic and doctrinal in nature. It does not constitute legal advice. The positions expressed are the author's own.