US-State
Voice Recording Consent by State
The short answer. Wiretap consent and dataset consent are different questions, and clearing the first does not clear the second. A state's recording law decides whether you were allowed to make the recording. What a data buyer needs is the speaker's agreement to the training and distribution uses, which comes from consent law and right-of-publicity law rather than from the wiretap statutes. The state that matters is where the speaker was, not where your company sits. Three states — California, New York and Tennessee — name voice expressly in their publicity statutes.
The law
18 U.S.C. §2511(2)(d)
It shall not be unlawful under this chapter for a person not acting under color of law to intercept a wire, oral, or electronic communication where such person is a party to the communication or where one of the parties to the communication has given prior consent to such interception unless such communication is intercepted for the purpose of committing any criminal or tortious act in violation of the Constitution or laws of the United States or of any State.
This provision answers the recording question and stops there. It says nothing about training a model on the recording, redistributing it, or using it in commerce.
Cal. Civ. Code §3344(a)
Any person who knowingly uses another's name, voice, signature, photograph, or likeness, in any manner, on or in products, merchandise, or goods, or for purposes of advertising or selling, or soliciting purchases of, products, merchandise, goods or services, without such person's prior consent, or, in the case of a minor, the prior consent of his parent or legal guardian, shall be liable for any damages sustained by the person or persons injured as a result thereof.
Voice is named in the statute, so California treats a voice as a protected attribute of identity rather than only as data. A synthetic voice built from a California speaker's recordings, or a dataset sold as a product, engages this section.
N.Y. Civ. Rights Law §51
Any person whose name, portrait, picture or voice is used within this state for advertising purposes or for the purposes of trade without the written consent first obtained as provided in section fifty of this article may maintain an equitable action in the supreme court of this state against the person, firm or corporation so using his name, portrait, picture or voice, to prevent and restrain the use thereof; and may also sue and recover damages for any injuries sustained by reason of such use.
New York also names voice, and it requires the consent to be in writing, so a verbal agreement will not satisfy it. The remedy is equitable as well as monetary.
Tenn. Code Ann. §47-25-1102
"Voice" means a sound in a medium that is readily identifiable and attributable to a particular individual, regardless of whether the sound contains the actual voice or a simulation of the voice of the individual.
Tennessee added this definition through the ELVIS Act, effective July 1, 2024, and the second half is what matters: a simulation counts as a voice, so a cloned voice model is covered even where no recording of the person is used.
Who it applies to
The rule that governs is where the speaker was, not where the buyer is. A dataset assembled from speakers in California, New York or Tennessee carries those states' publicity requirements into a contract signed in another country, because the wrong is committed where the person is protected rather than where the company operates. That is why the speaker's location belongs in the manifest.
The consent you need is speaker-level rather than session-level. A production company that organized the recording can grant you rights in its own work, but it cannot grant you the speaker's consent to being used in trade, and it cannot grant it on the speaker's behalf. Where a supplier offers "full rights" in a corpus, the question to ask is which of the two rights that phrase is meant to cover.
Right of publicity extends past death in many states, which matters for archival material. California provides a post-mortem right for a period of decades, and other states vary widely, with some recognizing none at all. Historical broadcast audio and estate-held archives therefore need an estate consent in addition to the speaker's, and that consent is often unavailable.
Children are treated more strictly across the board. A guardian's consent is required for publicity purposes in California and elsewhere, and the permitted use should be stated rather than implied. This is a category where the documentation burden is highest and where we require a stated permitted use before taking a project on.
- Commissioned recordings of consenting adults: the consent document can cover recording, training and distribution in one form.
- Agency or studio corpora: check whether the speaker consented to use in trade, not only to being recorded.
- Archival and estate material: needs a post-mortem publicity consent, which frequently does not exist.
What it costs to get wrong
California provides statutory damages under Civil Code §3344(a) of the greater of 750 dollars or actual damages, plus the defendant's profits attributable to the use, and attorney's fees. The 750 dollars floor matters because it is per use rather than per plaintiff, and a dataset distributed to multiple customers is not one use.
New York allows compensatory and, on the right facts, punitive damages under Civil Rights Law §51, alongside the injunction that the section makes available as of right. Tennessee provides a civil action and treats a violation as a Class A misdemeanour, which carries up to eleven months and twenty-nine days and a fine of up to 2,500 dollars, and the ELVIS Act extended liability to anyone distributing a tool whose primary purpose is producing a particular person's voice.
The consequence that decides deals is the injunction rather than the damages figure. A voice model or a dataset that has to be withdrawn is a product failure, and unlike a damages award it cannot be absorbed and continued past. Buyers who have shipped a synthetic voice built on unclear consent are exposed on every deployment, not once.
For datasets, the practical cost is often paid earlier and more quietly. A supplier who cannot show speaker-level consent for use in trade will fail a procurement review, and the deal ends before any of these statutes is ever cited.
How to comply when you are buying data
The distinction to keep hold of is that recording consent and use consent are two documents. Most suppliers have the first. Fewer have the second, and the second is the one a data buyer depends on.
The consent you need is two documents, not one, and suppliers often hold only the first. The recording release answers the wiretap question. The use consent answers the publicity and data protection question, and it is the one that names training, distribution and synthetic output. Asking for "the consent" without distinguishing the two is how buyers end up holding a document that clears the wrong thing.
Where a project involves speakers in states that name voice in a publicity statute, we collect the speaker-level consent to use in trade alongside the recording release, and we decline projects where only the production company can be shown to have agreed. We are a sourcing company rather than a law firm, so this is documentation of what was agreed rather than advice on what your jurisdiction requires.
- Ask for speaker-level consent, not just a supplier warranty. A warranty transfers the argument; it does not create the consent, and it is capped at the contract value.
- Check that the consent names the training and distribution uses. A release covering recording and internal use does not reach a model you ship to customers.
- Get the speaker state in the manifest, because that is what determines which publicity statute applies and whether a post-mortem right is in play.
- Check the synthetic voice question explicitly. Under Tennessee's definition a simulation is a voice, so the consent should say whether creating a synthetic version is permitted.
- Ask about post-mortem rights if the source is archival, and expect the answer to be a gap rather than a document. Estates rarely pre-authorize training use.
Related compliance topics
-
EU AI Act Transparency Requirements
eu ai act transparency requirements
-
EU AI Act Article 50
eu ai act article 50
-
Data Provenance
data provenance
-
Data Licensing Agreement
data licensing agreement
-
AI Training Data Laws
ai training data laws
-
AI Training Data Governance
ai training data governance
Not legal advice
We are a sourcing company, not a law firm. Nothing on this page is legal advice, and it does not create a lawyer–client relationship. Whether a particular dataset is permissible in your jurisdiction depends on your use case, where you operate, and where the people in the recordings are located. Our role is to document the chain of consent accurately so that your counsel can assess it.
Sourcing data under Voice Recording Consent by State?
Tell us the language, the hours, and what the data needs to look like. You will get a real number and a real timeline — not a range. If we cannot source it well, we will tell you that instead.
- Pilot batch before the full run, so problems surface early.
- Consent documentation delivered with the data.
- No medical or clinical data. No recorded telephone calls.