he biggest concerns voice-over talent have about AI tend to fall into five connected areas: ownership, employment, compensation, consent, and creative identity.

1. Voice cloning without meaningful consent

This is probably the biggest concern. A performer may provide recordings for one job and discover that those recordings—or a voice model built from them—can generate entirely new performances.

The concern isn’t simply “AI sounds like me.” It’s:

• Who is allowed to create the replica?
• What exactly can it be used for?
• For how long?
• Can the permission be revoked?
• Can the voice be sold or licensed to someone else?
• Can it be used for a product, character, political message, or context the actor would never personally endorse?

SAG-AFTRA’s current framework specifically emphasizes clear consent, fair compensation, and control over the performance, and recent contracts require more specific disclosure of intended uses. SSAG-AFTRA+1

2. Losing work to synthetic voices

There’s a very practical fear: clients won’t hire a human if an AI voice is cheaper, faster, and available 24/7.

This is particularly concerning for:

• e-learning
• corporate narration
• explainer videos
• audiobooks
• advertising
• video games
• localization/dubbing
• IVR/phone systems
• social-media content

Voice actors aren’t necessarily opposed to AI as a tool. The fear is that it changes the economics from “hire a performer” to “buy access to a voice.”

SAG-AFTRA has gone so far as to explicitly negotiate language recognizing that voice actors are human performers, while establishing compensation when digital replicas are used.

3. Being paid once for something that earns money indefinitely

This is a particularly important issue.

Imagine a voice actor is paid $500 for a recording session. That recording is then used to train or create a voice model that produces thousands of hours of narration over several years.

The actor’s concern becomes:

“Why am I being paid for one session when my digital performance is effectively doing thousands of future sessions?”

That’s why performers are pushing for usage-based compensation, royalties, residuals, or separate licensing fees rather than treating AI rights as an incidental part of a recording session.

SAG-AFTRA’s voice/animation agreements, for example, provide compensation mechanisms for the use of employment-based digital replicas and residuals for certain digitally altered foreign-language performances.

4. Contracts that quietly give away AI rights

This is a huge concern for working talent.

An actor may see language such as:

“The producer may use recordings, performances, likeness, or voice for any purpose…”

Historically, a clause like that might have had relatively limited implications. With modern generative AI, it could potentially mean something much more consequential.

Performers therefore worry about AI rights being buried in broad contract language, rather than explicitly negotiated.

SAG-AFTRA’s current guidance calls for consent to be clear and conspicuous, with a reasonably specific description of intended use—not simply hidden in fine print.

5. Losing control of their professional identity

A voice isn’t just an audio file. For many VO professionals, it’s their brand and livelihood.

Someone might build a recognizable commercial, character, narration, or celebrity-like voice. An AI replica could then say things the real performer would never say.

That raises questions such as:

• What if my AI voice endorses something I oppose?
• What if it says something offensive?
• What if it appears in pornography or political advertising?
• What if it is used to impersonate me?
• What if the synthetic voice becomes associated with poor-quality work?
• What if clients start thinking the AI voice is my actual performance?

The broader performer-rights movement is therefore increasingly framing voice as a form of identity and intellectual property, not merely a recording.

6. Training-data concerns

There’s another, somewhat different issue: how AI companies obtained the voices used to train their systems.

VO talent may ask:

“Was my work used to train an AI model without my permission?”

This gets into copyright, licensing, database scraping, and whether a recording that was legally available online can legitimately be used to train a commercial voice-generation system.

It’s one reason performer organizations have pushed for AI systems to respect permission and compensation rather than assuming that publicly accessible creative work is free training material.

7. The erosion of the craft itself

There’s also a less quantifiable concern: voice acting is more than producing intelligible words.

Human performers bring:

• acting choices
• timing
• emotional interpretation
• character development
• improvisation
• subtext
• interaction with directors
• cultural understanding
• intentional imperfections

A synthetic voice may reproduce the sound of a performance without reproducing the human creative process behind it.

For some actors, that’s a concern about their livelihood; for others, it’s a concern about what happens to the art form when clients increasingly optimize for cost and speed.

---

The interesting part: many VO artists aren’t simply “anti-AI”

This distinction is important.

A lot of the professional conversation has moved toward “AI under performer control” rather than “AI must never exist.”

For example, SAG-AFTRA’s current AI framework explicitly allows for arrangements in which performers voluntarily license their voices to AI companies, provided there is informed consent, compensation, and control.

So the fundamental position is increasingly:

AI as a tool: 👍 AI that a performer knowingly licenses: 👍 AI that replaces a performer without their permission: 👎 AI trained on or cloned from someone’s work without authorization: 👎 AI rights hidden in a contract: 👎 A one-time fee for unlimited exploitation: 👎

In other words, the core concern isn’t necessarily “AI can make a voice.” It’s “Who owns the right to make that voice, who controls it, and who gets paid when it generates value?”