Voice Cloning for Business Calls: Capabilities, Consent and Compliance
Voice cloning has crossed from novelty to production tool. A few minutes of recorded speech is now enough to build a synthetic voice that delivers outbound calls in a founder’s tone or a sales director’s accent — at any volume, in any language, around the clock. It is also exactly the technology regulators on both sides of the Pacific spent 2024 writing rules about. So the question is not just “can we?” but “under what consent, disclosure and identification obligations?”
This article is general information, not legal advice. Regulations change and enforcement positions shift — engage qualified counsel in each market you call before deploying cloned voices.
Can businesses legally use cloned voices on calls? Yes — with layered consent. The voice’s owner must consent to the cloning, and called parties must give the consent their jurisdiction requires. Under the FCC’s February 2024 ruling, AI-generated voices are “artificial” under the TCPA: US telemarketing calls need prior express written consent, identification and opt-out. Australian calls must follow Do Not Call and telemarketing-standard rules.
What voice cloning actually does on a business call
A cloned voice is a text-to-speech model trained on recordings of a specific person, so that anything the system says comes out in that person’s voice. On a live business call, the clone is the delivery layer: a conversational AI decides what to say, and the cloned voice says it — handling interruptions and objections in real time rather than playing a fixed recording.
Three capabilities matter commercially. Consistency: every call sounds the same on the thousandth dial as on the first, with no fatigue or turnover. Scale: one consented voice can hold many simultaneous conversations. Multilingual delivery: modern platforms can render the same consented voice across dozens of languages — Zian’s agents operate in 30+ languages across live phone, SMS, email and WhatsApp — so a spokesperson who only speaks English can greet a prospect in Spanish or Vietnamese in a voice that is recognisably theirs. For a broader look at conversational voice agents on outbound work, see our guide to the best AI voice agents for outbound calls in 2026.
None of this changes the legal character of the call: a cloned voice makes it more regulated, not less.
Legitimate uses versus prohibited uses
Legitimate use cases share a structure: a real person has knowingly licensed their voice, and the recipients have consented to the calls. Brand voice consistency — your own spokesperson’s voice on your own consented outreach — sits here, as does multilingual delivery of that consented voice. Answering inbound calls is also a materially lower-risk posture in the US, because the TCPA’s restrictions attach to calls a business initiates.
The prohibited end is equally clear. Cloning a voice you have no rights to — a celebrity, a competitor’s executive, a government official — is impersonation, and regulators treat it as fraud. The FTC finalised its Government and Business Impersonation Rule in February 2024 and simultaneously proposed extending it to impersonation of individuals, citing AI voice cloning explicitly. It also ran a public Voice Cloning Challenge (winners announced April 2024) and used that announcement to reaffirm that the Telemarketing Sales Rule applies to AI-enabled scam calls. The other bright line is unconsented robocalling — unlawful whether the voice is a crude robot or a perfect clone.
Between the two ends sits a grey zone worth naming honestly: relying on old “prerecorded voice” consents for new AI-generated calls, cloning the voice of an employee who has since resigned, or using a consented voice outside the licensed scope. These are the cases pending rulemaking is most likely to tighten.
| Use case | Status | Governing rule |
|---|---|---|
| Cloned spokesperson voice on outbound calls to contacts with prior express (written, for marketing) consent, identification and opt-out | Legitimate | TCPA / 47 CFR § 64.1200 (US); Do Not Call Register Act 2006 + Telemarketing Industry Standard 2017 (AU) |
| Multilingual delivery of a consented brand voice to opted-in lists | Legitimate | Same consent and identification rules per market and language |
| Cloned voice answering inbound calls | Generally lower risk | TCPA restrictions attach to outbound calls; the FCC’s 2024 NPRM notes the TCPA’s requirements “do not extend to technologies used to answer inbound calls” |
| Using pre-2024 “prerecorded voice” consents for AI-generated calls without fresh disclosure | Grey zone | FCC NPRM (FCC 24-84) proposes disclosure at the point of consent; not yet final as of July 2026 |
| Cloning an ex-employee’s voice, or exceeding the licence scope | Grey zone to prohibited | Contract and privacy law; consent scope and revocation terms |
| Impersonating a real person, business or government; unconsented cloned-voice robocalls | Prohibited | FTC Impersonation Rule and Telemarketing Sales Rule; TCPA via the FCC’s February 2024 Declaratory Ruling |
The US rules: the TCPA and two FCC actions
The foundational instrument is the Telephone Consumer Protection Act. On 8 February 2024 the FCC released a Declaratory Ruling (FCC 24-17, CG Docket No. 23-362) confirming that the TCPA’s restrictions on “artificial or prerecorded voice” calls encompass current AI technologies that generate human voices — naming voice cloning specifically, because “this technology artificially simulates a human voice”. The consequences are concrete:
- Prior express consent is required before calling with an AI-simulated or AI-generated voice, absent an emergency or exemption — and if the call introduces an advertisement or constitutes telemarketing, consent must be prior express written consent.
- Identification: artificial or prerecorded voice messages must state clearly, at the beginning, the identity of the business responsible for the call, and provide its telephone number during or at the end (47 CFR § 64.1200(b)(1)–(2)).
- Opt-out: messages that include or introduce an advertisement or constitute telemarketing must offer specified opt-out methods to stop future calls (47 CFR § 64.1200(b)(3)).
The ruling also closed a door some vendors leaned on: the TCPA “does not allow for any carve out of technologies that purport to provide the equivalent of a live agent”. A conversing clone is still an artificial voice in law.
The second action is still in flight. In August 2024 the FCC adopted a Notice of Proposed Rulemaking (FCC 24-84, published in the Federal Register in September 2024) proposing to define an “AI-generated call” as one that uses “any technology or tool to generate an artificial or prerecorded voice or a text” — expressly including machine learning and large language models — to communicate with a called party on an outbound call. It proposes two new disclosure duties: callers would have to disclose their intent to use AI-generated content when obtaining consent, and disclose the use of an AI-generated voice at the beginning of each call. Comments closed in October 2024, and as at 30 July 2026 the FCC has not adopted final rules in the docket — the AI-specific disclosures remain proposals, not law. The prudent read: the consent, identification and opt-out obligations already bind you today, and building call-start AI disclosure in now means the pending rules cost you nothing if finalised. Our companion piece on outreach compliance for AI agents maps these rules alongside the Spam Act and GDPR in more depth.
The Australian rules: Do Not Call, the telemarketing standard and the Spam Act
Australia has no voice-cloning-specific statute yet, so cloned-voice calls are governed by the general telemarketing framework — which is stricter than many US operators expect. The Do Not Call Register Act 2006 prohibits unsolicited telemarketing calls to numbers on the register (section 11), so lists must be washed before dialling. Layered on top, the Telecommunications (Telemarketing and Research Calls) Industry Standard 2017 applies to telemarketing calls to Australian numbers regardless of registration status: calls are permitted 9:00 am–8:00 pm on weekdays and 9:00 am–5:00 pm on Saturdays, with none on Sundays or national public holidays; callers must enable calling line identification and keep the return number reachable for at least 30 days; the caller’s identity and the call’s purpose must be provided; and the call must be terminated immediately if the recipient asks. Every one of those is a behaviour a cloned-voice agent must be configured to exhibit, not a policy document.
The Spam Act 2003 sits beside this framework but covers a different channel: commercial electronic messages — email, SMS and instant messaging — where it requires consent, accurate sender identification and a functional unsubscribe (sections 16–18). A cloned-voice phone call is not an electronic message under that Act, but the moment your campaign follows a call with an SMS or WhatsApp message, you are inside it. Voice recordings of identifiable people are also personal information under Australian privacy law, so the voice you clone needs its own privacy footing.
A practical consent-and-disclosure checklist
Two consent relationships must both be in place before a cloned voice dials anyone.
The voice owner:
- A written licence from the person whose voice is cloned, naming the permitted uses, channels, markets and languages.
- A defined term and a revocation path — what happens to the model and pending campaigns if they withdraw or leave.
- Secure storage of the training recordings and voice model, treated as personal information.
The called party:
- Prior express consent on record for artificial-voice calls — written for anything that markets in the US — captured with provenance: who, when, how, to what.
- Disclosure at the point of consent that calls may use an AI-generated voice (proposed US law; good practice everywhere now).
- Identification at the start of every call: the business responsible, a working callback number — and, as emerging best practice, that the voice is AI-generated.
- A functioning opt-out honoured immediately and propagated across every channel.
- List washing against the Do Not Call Register for Australian numbers, and calling windows enforced in the recipient’s local time.
- Full call recordings and logs retained as compliance evidence.
How a responsible platform implements this
The honest position for any vendor is that voice cloning ships inside a compliance envelope, or it should not ship at all. At Zian, voice cloning is a supported capability of agents that work across live phone, SMS, email and WhatsApp — and the platform treats the rules above as configuration, not customer homework. SmartReach AI™ orchestrates message, channel and timing — enforcing calling windows and channel-appropriate consent instead of leaving them to operator memory. PrecisionPitch AI™ continuously split-tests how agents open and handle objections — with identification and disclosure lines held fixed, because you optimise conversion inside the rules, never by testing your way out of them. Consent provenance and opt-outs live in your CRM of record through HubSpot, Salesforce, HighLevel and Zapier integrations, so an opt-out on one channel stops every channel. For organisations whose voice models and call data cannot leave their environment, private model deployment is available. Compliance and performance are not in tension — the AI books 40+ meetings/week for many teams, on lists those teams are entitled to call. If you are weighing this against traditional operations, our comparison of AI voice agents vs call centres covers the auditability difference. To see the consent-first architecture on your own pipeline, Apply For Partnership.
Frequently asked questions
Is it legal to use a cloned voice on sales calls in the United States?
Yes, with consent. The FCC’s February 2024 Declaratory Ruling confirms that AI-generated and cloned voices are “artificial” voices under the TCPA, so a sales call using one requires the called party’s prior express consent — prior express written consent if it contains marketing — plus caller identification at the start of the message and an opt-out mechanism for telemarketing messages.
Do I have to tell people they are talking to an AI?
In the US you must identify the business responsible at the beginning of any artificial-voice message, but a specific “this voice is AI-generated” disclosure is currently only a proposal: the FCC’s 2024 rulemaking would require it at the point of consent and at the start of each call, and as at 30 July 2026 those rules have not been finalised. In Australia, the telemarketing standard requires you to identify the caller and the call’s purpose. Disclosing the AI voice anyway is cheap, honest and future-proof.
Whose consent do I need before cloning a voice?
Two layers. First, the person whose voice you clone: a written licence covering the uses, channels, markets and languages, with a revocation path. Second, the people you call: they must have consented to receive artificial-voice calls under the rules of their jurisdiction. Consent from one layer never substitutes for the other — a perfectly licensed voice on an unconsented call list is still an unlawful campaign.
Does Australia’s Spam Act cover cloned-voice phone calls?
No. The Spam Act 2003 covers commercial electronic messages — email, SMS and instant messaging — where it requires consent, sender identification and a working unsubscribe. Voice calls are instead governed by the Do Not Call Register Act 2006 and the Telemarketing and Research Calls Industry Standard 2017, which restrict calling to registered numbers, set permitted calling hours, and require caller identification and immediate termination on request.
Can I clone a celebrity’s or competitor’s voice for marketing?
No. Using a voice you have no rights to is impersonation. The FTC finalised its Government and Business Impersonation Rule in February 2024 and proposed extending protection to impersonation of individuals, citing AI voice cloning as a driver, and the Telemarketing Sales Rule applies to AI-enabled scam calls. Only clone voices you hold a written licence for.
What is the status of the FCC’s proposed AI-call disclosure rules?
Still pending. The FCC adopted a Notice of Proposed Rulemaking (FCC 24-84) in August 2024 proposing a definition of “AI-generated call” and disclosure duties at the point of consent and at the start of each call. Comments closed in October 2024, and as at 30 July 2026 no final rules have been adopted in the docket. The February 2024 Declaratory Ruling, by contrast, is already in force.
Voice cloning on business calls is neither a loophole nor a landmine — it is a regulated capability that rewards operators who build consent and disclosure in from day one. If that is the outbound operation you want to run, Apply For Partnership.