AI Resurrections, Synthetic Endorsements, and the Consent Problem That Doesn’t Die
The dead are working again.
AI-generated Elvis is performing in a London immersive experience. A synthetic James Earl Jones is voicing Darth Vader in Fortnite. An AI-generated George Carlin delivered a comedy special he never wrote, to an audience he never met, about a world he never lived in.
Meanwhile, OpenAI’s Sora has made it trivially easy for anyone to generate photorealistic video of deceased celebrities—saying things they never said, selling things they never endorsed, appearing in contexts their estates never approved.
For executives, this isn’t a Hollywood problem. It’s a governance problem wearing a famous face. And if your organization uses AI to generate content—marketing, customer-facing, internal—you’re closer to this than you think.
What’s Happening: The New Economics of the Dead
The technology to digitally resurrect deceased individuals has existed in crude form for over a decade—the Tupac hologram appeared at Coachella in 2012. What’s changed is the cost curve and the access curve. Generating a convincing synthetic likeness used to require a studio’s VFX budget. Now it requires a text prompt.
That shift turns a niche entertainment question into an enterprise-wide risk. The cases that matter aren’t just about dead celebrities—they’re about the governance principles that apply to anyone whose likeness, voice, or identity could be synthesized. Consider three recent examples:
The George Carlin Special. In January 2024, a podcast company released “George Carlin: I’m Glad I’m Dead,” an AI-generated comedy special that mimicked the late comedian’s voice and style to produce new material he never performed. Carlin’s estate sued immediately. The special was taken down. The estate’s position was unambiguous: no one had permission to use Carlin’s likeness, and no machine can replace the creative intent of the artist. The producers claimed they were paying “homage.” The court didn’t need to rule—the lawsuit settled—but the reputational damage to the producers was total.
James Earl Jones and Fortnite. In 2025, SAG-AFTRA filed an unfair labor practice charge against the producers of Fortnite for using an AI-generated version of the late James Earl Jones’s voice as Darth Vader—without bargaining with the union. The union argued the practice not only violated the rights of the deceased performer but also displaced work that living performers could have done. This case moved the issue from “estates and heirs” into labor and employment law.
The ABBA Voyage Model. On the other end of the spectrum, ABBA’s Voyage virtual concert—featuring digital “ABBAtars” of the still-living band members—became a commercial and critical success. The key difference: the artists controlled the process, approved their representations, and were central to the creative decisions. A 2026 study from the University of South Florida found that audience acceptance of AI hologram concerts depended primarily on ethical considerations—respect for the artist’s legacy and transparency—rather than on the technological novelty itself.
Three cases. One pattern: consent and control are the dividing line between innovation and exploitation.
Why This Is an Enterprise Problem, Not a Hollywood Problem
It’s tempting to file this under “entertainment industry drama.” That’s a mistake.
The same technology that resurrects Elvis can synthesize your former CEO for an internal video, generate a deceased founder’s voice for a brand campaign, or produce a likeness of a retired spokesperson who never agreed to AI use in their original contract. Every organization that creates AI-generated content—visual, audio, or text—is operating in the same legal and ethical territory as the studios.
And the legal ground is shifting fast.
California’s AB 1836 and AB 2602, which took effect in early 2025, now prohibit the use of digital replicas of deceased performers without estate consent and void contract provisions that allow AI replacement of living performers without informed, specifically described consent. Tennessee’s ELVIS Act provides similar protections. Illinois amended its Right of Publicity Act in August 2024 (HB 4875, effective January 2025) to prohibit unauthorized digital replicas. New York enacted two laws in December 2025: one expanding posthumous publicity rights to cover digital replicas (effective immediately), and a separate first-in-the-nation synthetic performer disclosure law requiring advertisers to identify AI-generated performers (effective June 2026).
At the federal level, the NO FAKES Act—first introduced in 2023, formally proposed in 2024, and reintroduced in April 2025—would create a national right over digital replicas for both living and deceased individuals. It hasn’t passed yet, but the direction is clear.
For executives, the operative reality is a patchwork. There is no single federal standard. State laws vary in scope, duration, and enforcement mechanism. And the gap between what is technically possible and what is legally permitted is widening faster than legislatures can close it.
The Consent Spectrum: A Framework for AI Likeness Risk
Not all synthetic likenesses carry the same risk. The problem isn’t AI-generated content itself—it’s the governance architecture around it. I’d propose that every AI likeness decision sits somewhere on a consent spectrum with three distinct zones:
Zone 1: Authorized Resurrection
The estate or individual explicitly consents. The representation is reviewed, controlled, and approved. Creative direction remains with the rights holder. ABBA Voyage is the benchmark. So is Peter Cushing’s appearance in Rogue One, which was negotiated with his estate. Even here, risk remains: consumer backlash can still follow if audiences feel the use is exploitative, regardless of legal authorization. The 2025 Elvis Evolution immersive show in London received mixed reactions despite estate involvement—some audiences felt the production oversold what it delivered.
Zone 2: Contractual Gray Area
A contract exists, but it was written before generative AI made synthetic performances possible. The language grants broad “likeness rights” or “image use” without specifying AI replication. This is where the James Earl Jones case lives—and it’s where most legacy talent agreements sit. During the 2023 Hollywood strikes, background actors reported that their body scans were being stored for potential AI reuse, sometimes described as being available “for the rest of eternity,” without clear terms or additional compensation. California’s AB 2602 was a direct legislative response: it voids contract provisions that allow digital replacement without informed consent and proper representation.
Zone 3: Unauthorized Replication
No consent. No contract. Someone generates a synthetic likeness because they can. The George Carlin special lived here. So do the Sora-generated videos of deceased celebrities circulating on social media. This is the fastest-growing zone and the hardest to enforce against, precisely because the tools are now consumer-grade. The liability exposure isn’t just legal—it’s reputational. A brand that unknowingly uses AI-generated content trained on unauthorized likenesses inherits the risk.
The uncomfortable insight is that even Zone 1 isn’t safe. Authorization from the estate doesn’t guarantee consumer acceptance. And consumer acceptance today doesn’t prevent regulatory action tomorrow. Research consistently shows that audiences care less about technical sophistication and more about whether the use feels respectful. That’s a judgment call, not a contract clause—and it’s exactly the kind of decision that organizations are worst at making under commercial pressure.
What Senior Leaders Should Do Now
1) Audit every contract that touches likeness, voice, or image rights.
If your talent agreements, spokesperson contracts, or vendor content licenses were drafted before 2023, they almost certainly don’t address AI-generated replicas. This isn’t hypothetical—California’s AB 2602 now voids ambiguous provisions. A contract that says “all media now known or later developed” is no longer a reliable basis for synthetic content.
2) Establish a synthetic content policy before your marketing team needs one.
Define what AI-generated likenesses your organization will and won’t use. This includes likenesses of: deceased individuals (founders, brand figures, historical references), current or former employees and spokespersons, public figures and celebrities, and composite or fictional personas that may resemble real people. The policy should cover not just marketing, but training data, internal communications, and any customer-facing AI that generates visual or audio output.
3) Add AI likeness provisions to your procurement and vendor contracts.
If you work with creative agencies, content platforms, or AI vendors that generate visual or audio assets, your contracts need to address: provenance (can the vendor certify the training data didn’t include unauthorized likenesses?), indemnification (who bears the cost if a synthetic likeness generates a legal claim?), and disclosure (will AI-generated likenesses be labeled as such in the final deliverable?). If your agency can’t answer these questions, your exposure is real.
4) Decide on transparency before the market decides for you.
Research from a Getty Images–commissioned survey found that nearly 9 in 10 consumers globally want to know whether an image was created using AI, and 98% said authentic images are critical for establishing trust. A 2025 study from the Nuremberg Institute for Market Decisions found that simply labeling an ad as AI-generated reduced perceived naturalness and purchase willingness. New York’s synthetic performer disclosure law, signed in December 2025 and effective June 2026, will require advertisers to identify AI-generated performers—the first state law of its kind. The regulatory trajectory points one direction: mandatory transparency. Leaders who adopt it voluntarily now build the trust infrastructure before it’s imposed.
5) Map your jurisdictional exposure.
California, Tennessee, Illinois, and New York have enacted AI likeness laws, with New York’s disclosure requirement taking effect in June 2026. The federal NO FAKES Act remains in committee. If your organization creates or distributes AI-generated content across state lines—and in 2026, nearly everyone does—you need a compliance map, not a single-jurisdiction opinion.
The Deeper Question
Underneath the contract language and the compliance mapping, there’s a question worth sitting with: what does it mean to consent to something you can’t yet imagine?
James Dean died in 1955. He couldn’t have consented to a CGI performance in a Vietnam War film because the concept didn’t exist. George Carlin couldn’t have anticipated an AI system trained on his material producing new routines. The background actors who had their bodies scanned in 2022 couldn’t have understood what “all media later developed” would mean two years later.
This is the consent paradox of generative AI: the technology invalidates the assumptions under which consent was originally given. Every talent agreement signed before the generative AI inflection point is, to some degree, an artifact of a world that no longer exists. And every new agreement must somehow anticipate uses that don’t yet exist.
For executives, this isn’t a philosophical exercise. It’s an operational reality. The organizations that treat consent as a one-time checkbox—rather than an ongoing obligation—are the ones that will find themselves in litigation, in reputational crisis, or both.
North Star
The dead can’t renegotiate their contracts. That’s exactly why the living have to get the governance right.
If your AI content strategy can’t survive the question “Did the person depicted—or their estate—specifically authorize this use, in this form, for this purpose?” then it isn’t a strategy. It’s a liability with a launch date.
Executive Checklist
- Have we audited every talent, spokesperson, and content license agreement for AI-specific likeness provisions?
- Do we have a written synthetic content policy that covers marketing, training data, internal comms, and customer-facing AI?
- Can our creative agencies and AI vendors certify the provenance of likenesses in their outputs?
- Do our vendor contracts include indemnification for unauthorized likeness claims?
- Have we adopted a disclosure standard for AI-generated visual and audio content?
- Do we know which state AI likeness laws apply to our content distribution footprint?
- Is there a designated owner for synthetic content governance—not just legal, but cross-functional?
- If we discovered tomorrow that our AI vendor used unauthorized training data, do we have a response plan?
Sources
California Assembly Bill 1836 (2024). Signed into law September 17, 2024, effective January 1, 2025. Prohibits the use of digital replicas of deceased personalities without estate consent.
California Assembly Bill 2602 (2024). Signed into law September 17, 2024, effective January 1, 2025. Voids contract provisions allowing AI replacement of performers without informed consent.
Dogan, S. et al. (2026). “Reviving legends through holographic AI event experiences: Consumer acceptance and value insights.” International Journal of Contemporary Hospitality Management.
George Carlin Estate v. Dudesy et al. (2024). Lawsuit filed January 2024, Los Angeles Superior Court. Challenged AI-generated comedy special mimicking Carlin’s voice and style.
Getty Images / Prodoto (2024). “Imagery in the wake of AI.” Consumer survey on AI-generated imagery, authenticity, and trust in visual media. Note: Survey was commissioned by Getty Images. https://www.prodoto.com/behind-prodoto/blog/imagery-in-the-wake-of-ai
Illinois House Bill 4875 (2024). Signed into law August 9, 2024, effective January 1, 2025. Amends the Illinois Right of Publicity Act to prohibit unauthorized AI-generated digital replicas.
New York S. 8420 / Synthetic Performer Disclosure Law (2025). Signed December 11, 2025, effective June 9, 2026. Requires conspicuous disclosure when advertisements use AI-generated synthetic performers.
New York Posthumous Right of Publicity Expansion (2025). Signed December 2025, effective immediately. Expands right of publicity for deceased personalities to cover digital replicas for commercial purposes.
NO FAKES Act of 2025 (S. 1367 / H.R. 2794). Reintroduced April 2025. Proposed federal right over digital replicas of voice and visual likeness.
Nuremberg Institute for Market Decisions (2025). Study on consumer perception of AI-labeled advertising content.
SAG-AFTRA (2025). Unfair labor practice charge filed against Llama Productions regarding AI-generated voice of James Earl Jones in Fortnite.
Tennessee ELVIS Act (2024). Ensuring Likeness, Voice, and Image Security Act. Prohibits unauthorized distribution of digital replicas.
