Technology | IP Law | Global Trends

In a pivotal ruling addressing the growing friction between artificial intelligence and creative rights, a federal judge in the Southern District of New York has allowed portions of a class action lawsuit by two professional voice actors to proceed against AI voice-generation startup Lovo Inc.

The case—Lehrman v. Lovo Inc.—raises timely and thorny questions at the intersection of AI training practices, publicity rights, and the extent of copyright protections in the digital age. U.S. District Judge J. Paul Oetken’s decision to uphold certain claims could have profound implications for how AI developers interact with human-created content—particularly voice recordings.

Case Overview: When AI “Steals” a Voice

Plaintiffs Paul Skye Lehrman and Linnea Sage, both working voice actors, allege that Lovo misused their voice recordings, initially submitted for limited purposes through freelance platforms like Fiverr. Those samples, the lawsuit claims, were then used without consent to create commercial AI-generated voices—branded under pseudonyms such as “Kyle Snow” and “Sally Coleman.”

The actors argue this unauthorized replication and sale of their vocal likenesses violates New York’s right of publicity statute and infringes upon their copyrights, while also constituting deceptive business practices.

Key Ruling Highlights

In a ruling issued July 10, 2025, Judge Oetken held that:

Claims That Survive

• Right of Publicity (New York Civil Rights Law § 50/51):
  The court found sufficient grounds for the plaintiffs to claim that the use of AI-generated voices resembling their own, when used in advertising and sale, may amount to a violation of their statutory right of publicity.
• Amended Copyright Claims Permitted:
  Although the court dismissed the original copyright claims, it granted plaintiffs leave to amend, allowing a more tailored argument that Lovo’s use of their actual recordings as AI training data constitutes infringement.

Claims Dismissed

• Trademark and False Advertising:
  The judge dismissed trademark claims on the basis that a voice, without a recognizable mark or name attached, doesn’t fall under Lanham Act protection.
• Mimicry ≠ Copyright Infringement:
  The court emphasized that the mimicry of a person’s voice—without copying the actual recordings—is not enough to establish copyright violation under existing law.

Legal Framework: Publicity Rights vs. Generative AI

This ruling is one of the first in federal court to explicitly consider whether AI-generated vocal likenesses can infringe on the right of publicity, a state-level protection that safeguards individuals’ names, images, and voices from unauthorized commercial use.

While copyright protects fixed expressions (like audio recordings), publicity rights extend to the unique characteristics of a person’s identity. In this case, Judge Oetken acknowledged that an AI-generated voice, if convincingly similar to a real person’s, could implicate publicity rights—especially if used for commercial gain.

Notably, New York’s law does not require the individual to be a celebrity, making it a powerful tool for everyday creative professionals impacted by unauthorized digital replicas.

The Implications for AI Developers & the Gig Economy

The lawsuit underscores a major legal blind spot in the fast-moving world of AI: the training of models on data scraped or sourced—sometimes contractually—through gig platforms.

If the amended copyright claim survives, it could:

• Force AI companies to seek explicit training-use licenses for data sourced from individuals.
• Establish boundaries between “input” and “output” rights in generative systems.
• Encourage reforms in gig economy terms of service to better protect creators.

For freelance creatives—particularly voice actors, audiobook narrators, and podcasters—the decision affirms growing legal recourse in defending their identity against algorithmic exploitation.

What Comes Next

The plaintiffs will now file an amended complaint, focusing on the copyright infringement claim related to AI training. Discovery is expected to explore:

• How Lovo obtained and processed the original recordings.
• Internal documentation about the use of Lehrman and Sage’s voices.
• How closely the generated voices resemble their originals.

Depending on the outcome, this could set precedent for future claims against AI companies operating in adjacent fields—such as music synthesis, deepfake video generation, and synthetic podcasting.

Conclusion: A Voice for the Voiceless?

The Lehrman ruling may be limited in scope, but its resonance is broad. At a time when artificial intelligence threatens to blur the line between mimicry and misappropriation, this case signals a growing judicial willingness to scrutinize AI developers’ data pipelines.

For now, voice actors have found a forum—and a federal judge willing to listen.