AI Transparency & Identity Rights — Marketing Compliance Scenario
We Already Have Hours of the Voice Artist’s Recordings on File. Can We Use AI to Clone Her Voice for the New Campaign Instead of Hiring Her Again?
A real AI transparency and intellectual property compliance scenario — with three decision options and the right answer.
Quick Answer
Can a company use existing recordings of a voice artist to train an AI and clone her voice for new work? No, not without explicit written permission for that specific use. The contract for the original recordings grants rights to those recordings, not to the voice itself. A voice artist owns their voice the same way a photographer owns their photographs. Cloning it for new work requires a separate license, regardless of what has already been paid for.
The Situation
Your marketing team is planning a new brand campaign. The voice artist from your last campaign brought the messaging to life — clients loved her delivery, and her voice has become associated with your brand. Hiring her again will cost $8,000. A team member points out that you have hours of her existing recordings on file from the previous project and suggests feeding them into an AI voice generator to recreate her voice for the new campaign lines. The technology is available, the recordings are in your files, and you already have a contract with her. The team lead needs to make a call.
What Should You Do?
Choice A Proceed with the AI voice generation. You already have an existing contract with the voice artist and the recordings are in your files. You have paid for the recordings — using them to create similar audio for a campaign is a natural extension of that relationship.
Choice B Contact the voice artist to let her know the plan — not to request permission, but as a professional courtesy. If she objects, you can reconsider. If she is fine with it, proceeding seems reasonable.
Choice C Do not proceed without explicit, written permission for this specific use. Approach the voice artist to discuss whether she is willing to licence her voice for AI-generated use, under what terms, and at what rate — or choose a different approved narrator for the new campaign.
The Right Call
Choice C — Obtain explicit written permission for this specific use, or choose a different narrator.
Choice B addresses the professional courtesy but not the legal requirement. Notifying someone that you plan to use their identity without their permission and giving them the opportunity to object is not the same as obtaining their consent. The contract grants rights to specific recordings, not to cloning the voice for new content. That is a separate license requiring a separate, written agreement.
The Recognition Insight
The contract gives you rights to the recordings — not to the voice itself. This distinction is the core of the compliance issue and the one most teams don’t see until it is explained. Paying for a photographer’s work gives you the rights to the photographs taken under that agreement. It does not give you the right to feed those photographs into an AI system to generate new images in that photographer’s style. Voice rights work the same way.
Why This Scenario Is Harder Than It Looks
The existing contract and existing recordings make it feel like you already own the right.
The team has paid for the voice artist’s work. The recordings are in their files. The contract is signed. All of these facts point to a legitimate, completed transaction — which is why using those assets further feels like a natural extension rather than a new activity requiring new permission. The AI step is where the transition from “using what we paid for” to “using something we didn’t pay for” happens invisibly.
The cost saving feels like smart business rather than a shortcut.
Eight thousand dollars is a real budget line. Finding a way to achieve the same result for less is good commercial thinking. The compliance issue is not the cost consideration — it is the assumption that the technology makes the action legitimate. AI making something technically possible does not make it legally or ethically permitted. The question is always whether you have the right, not whether you have the capability.
EU and UK regulatory frameworks explicitly take this position.
Under the EU AI Act and UK AI principles, using AI to impersonate or recreate a real person’s voice, image, or likeness without documented permission is impermissible. This is not an emerging or ambiguous area — it is a clearly stated obligation. The transparency principle that underpins both frameworks is that AI must not be used in ways that deceive people about who they are hearing or seeing. Creating new audio in a real person’s voice without their knowledge does exactly that.
Frequently Asked Questions
Does this only apply to professional voice artists, or to any person’s voice?
The right to one’s voice applies to any identifiable individual — professional voice artists, executives, employees, public figures, or private individuals. Cloning anyone’s voice without their explicit, documented permission for that specific use is an identity rights violation. The professional context of the voice artist in this scenario makes the contract dimension visible, but the underlying right is universal.
What would proper permission for AI voice use look like?
Explicit written agreement that specifies: the individual consents to AI cloning of their voice; the permitted use cases (specific campaign, product category, time period); whether the resulting AI audio can be used without disclosure that it is AI-generated; and the compensation for this specific additional use. Many voice artists are open to discussing AI voice licences — but it must be a separate, documented agreement, not an extension assumed from a prior contract.
What if we use AI to generate a voice that sounds similar but isn’t a specific person?
Generating a new AI voice that has no connection to a real individual is a different situation — it does not involve identity rights in the same way. However, it may still require disclosure under some regulatory frameworks if used in customer-facing content, and the resulting voice may have its own IP considerations depending on the AI tool’s terms of service. Consult your Legal team before using any AI-generated voice in external communications.
What training covers AI voice cloning and identity rights for marketing teams?
This scenario is part of the Responsible AI compliance training program from Xcelus, within the Transparency Rule section. The training covers identity rights, voice and likeness ownership, disclosure obligations, and the regulatory frameworks that make these requirements binding — including EU AI Act, UK AI principles, and emerging US regulation.
How to Use This Scenario in Training
AI policy and intellectual property training establish the principle. This scenario makes voice and identity rights concrete for marketing and creative teams who are actively encountering this decision.
Xcelus recommends this scenario for Marketing, Communications, Creative, and Production teams. The recognition skill is identifying the distinction between rights to a recording and rights to the voice, and understanding that AI making something technically possible does not make it legally permissible.
More Compliance Scenarios
|
We used AI to write sixty pieces of marketing content this quarter. Does the company own it? |
Can I paste the strategy meeting transcript into ChatGPT to save time on a summary? |
Browse all five Responsible AI compliance training scenarios. |
Want the Full Responsible AI Training?
Xcelus builds AI compliance training that covers the Transparency Rule — including voice and identity rights, disclosure obligations, copyright ownership of AI content, and the regulatory frameworks that make these requirements binding for your organization.