AI Voices on Phone Calls: Disclosure, Consent, and the Law

AI voice technology has moved quickly from experiment to everyday use. Businesses now use AI voice agents to answer calls, make reminders, follow up on inquiries, and handle parts of sales and support conversations. At the same time, AI-generated voices have been used to impersonate public figures, defraud families, and mislead voters.

Laws and regulators have been working to keep up. This article explains the main legal questions around AI voices on phone calls, particularly two that matter most: when businesses need consent to use AI voices, and when they need to disclose that a caller is an AI.

This article is for general information only and isn't legal advice. Laws in this area are developing quickly, so businesses should seek current legal advice for the places they operate and call.

Why AI Voices Raise New Questions

Earlier forms of phone automation were easy to recognize. A prerecorded message sounds like a recording, and a phone menu asks you to press a number. AI voices are different. They can sound natural, respond to what people say, and hold a full conversation. Many people can't reliably tell them apart from a human.

That creates two concerns. The first is intrusion: AI makes it possible to place enormous numbers of conversational calls cheaply, which could multiply unwanted calls. The second is deception: if people don't realize they're speaking to an AI, or believe they're speaking to a specific real person, they may be misled about who they're dealing with and what they're agreeing to.

Laws on AI voices broadly address these two concerns.

The United States

AI voices count as "artificial" under the TCPA

The most important development came in February 2024, when the Federal Communications Commission (FCC) unanimously ruled that calls using AI-generated voices count as calls using an "artificial or prerecorded voice" under the Telephone Consumer Protection Act (TCPA).

The ruling followed shortly after a high-profile incident in New Hampshire, where voters received robocalls using an AI imitation of President Biden's voice before the state's primary election.

The ruling means AI voice calls are subject to the same rules as prerecorded calls, including:

Consent. Non-marketing calls to mobile phones using an AI voice generally require the person's prior express consent, unless an exemption applies. Telemarketing calls using an AI voice generally require prior express written consent, for both mobile phones and residential landlines.

Identification. Calls using an artificial voice must identify the business responsible for the call at the beginning of the message and provide a phone number.

Opt-out. Telemarketing calls using an artificial voice must give people an automated way to opt out of future calls during the call.

In other words, a business can't avoid the TCPA's rules on prerecorded calls simply because its AI voice can hold a conversation.

Further FCC proposals

The FCC has also proposed additional rules specifically on AI-generated calls, including defining what counts as an AI-generated call and requiring businesses to disclose AI use when seeking consent and at the start of calls. Businesses should check whether any such proposals have since been adopted.

The Federal Trade Commission

The Federal Trade Commission (FTC) enforces laws against deceptive and unfair business practices, which apply regardless of the technology used. Using an AI voice to mislead people, for example by pretending to be a real person, a government agency, or another business, can break these laws.

In 2024, the FTC adopted a rule specifically prohibiting the impersonation of government agencies and businesses, giving it stronger powers to act against impersonation scams, including those using AI-generated voices. The FTC has also brought enforcement actions against companies making false or misleading claims involving AI.

State laws

Several states have passed laws addressing AI interactions, and many more have introduced proposals. Examples include:

Utah. Utah's Artificial Intelligence Policy Act, which took effect in 2024, requires businesses to disclose that a person is interacting with generative AI in certain situations, including when a consumer clearly asks whether they're speaking with a human, with stricter requirements for some regulated professions. The law has since been amended to clarify its scope.

Colorado. Colorado's AI Act includes a requirement for businesses that use AI systems to interact with consumers to tell them they're interacting with AI, unless it would be obvious. Its effective date was delayed to 2026, and businesses should check its current status.

California. California has had a law since 2019 that makes it unlawful to use a bot to communicate with people online with the intent to mislead them about its artificial identity, in order to encourage a sale or influence a vote.

Voice and likeness laws. Some states protect people's voices from unauthorized use. Tennessee's ELVIS Act, passed in 2024, specifically protects individuals against unauthorized use of their voice, including AI-generated imitations. Many states also have right of publicity laws protecting people's identities from unauthorized commercial use.

Election laws. A growing number of states restrict or require disclosure of AI-generated content, including audio, in election communications.

Because state laws differ and change frequently, businesses using AI voices across the US need to track the rules in each state they call.

The European Union

The EU's AI Act, which entered into force in 2024 with its obligations being phased in over several years, includes transparency obligations for AI systems that interact with people. In general, AI systems designed to interact directly with people must be built so that people are informed they're dealing with an AI, unless this is obvious from the context. The Act also requires AI-generated or manipulated audio, video, and images that could be mistaken for real, often called deepfakes, to be disclosed as artificially generated in many circumstances.

Alongside the AI Act, the GDPR applies to the personal data processed during AI calls, such as recordings, transcripts, and any decisions made about people. National rules on electronic marketing, based on the EU's ePrivacy rules, generally require consent for automated marketing calls.

The United Kingdom

The UK doesn't currently have a single law specifically governing AI. Instead, existing rules apply to AI voice calls:

PECR. Under the Privacy and Electronic Communications Regulations, automated marketing calls, which include calls using recorded or synthesized voices, require the person's specific prior consent to receive such calls from that organization.

Data protection. The UK GDPR requires businesses to be transparent about how they use people's personal data, including where AI is involved, and gives people rights in relation to automated decision-making that significantly affects them.

Consumer protection law. Misleading people about who or what they're dealing with can amount to an unfair or misleading commercial practice under UK consumer protection law.

Sector rules. Regulated sectors, such as financial services, have their own conduct rules that apply however a business communicates with customers.

Disclosure: Telling People They're Speaking With an AI

Disclosure requirements vary by place, but the direction of travel is clear. Being open about AI is increasingly expected by regulators, required in a growing number of places, and widely seen as the honest approach regardless of legal requirements.

Good disclosure practice generally means:

Disclosing early. Telling people at the very start of the call, ideally within the first few seconds, that they're speaking with an AI, before the conversation develops.

Using plain language. A simple statement like "I'm an AI assistant calling on behalf of [business]" is clearer than vague wording like "virtual representative."

Never denying it. If someone asks whether they're speaking with a real person, the answer should always be truthful.

Not impersonating real people. An AI should not claim to be a specific real person, and a cloned voice should not be used to suggest a real person is speaking when they aren't.

Disclosing changes. When a call moves from an AI to a person, or the other way around, telling the caller helps them understand who they're now speaking with.

Keeping the wording fixed. Many businesses make the disclosure a fixed line in the script, spoken word for word on every call, rather than leaving it to the AI to phrase.

Recording it. Keeping a record showing the disclosure was made on each call helps demonstrate compliance.

Consent: Getting Permission for AI Calls

Where consent is required, it needs to be meaningful. In practice, that means:

• Being specific about AI or artificial voices in consent wording, so people know the calls they're agreeing to may use automated or AI voices.
• Naming the business that will make the calls.
• Separating types of consent, such as live calls, automated calls, and texts.
• Not making consent a condition of purchase where the rules prohibit it.
• Keeping clear records of when, where, and how consent was given, and exactly what the person saw or heard.
• Honoring withdrawals quickly, through any reasonable method the person uses.

Inbound Calls

Most of the strictest rules focus on outbound calls, where the business contacts the person. When a person calls a business and reaches an AI voice agent, consent rules for automated calls usually don't apply in the same way, since the person chose to call.

However, transparency still matters. Many disclosure rules apply to AI interactions generally, not just outbound calls, and call recording rules, data protection law, and consumer protection law all apply to inbound calls too. Telling inbound callers they're speaking with an AI, and offering a route to a person, is good practice.

Preventing Misuse

Responsible use of AI voices also means guarding against misuse. Businesses and providers increasingly put safeguards in place, such as reviewing scripts for deceptive content before calls are made, preventing AI from impersonating government agencies or other organizations, requiring documented consent before cloning anyone's voice, restricting what AI agents are allowed to promise or claim, and keeping records of every call.

A Fast-Moving Area

AI voice regulation is changing rapidly. New laws, rules, and court decisions appear regularly, and proposals that are under discussion today may become requirements soon. Businesses using AI voices should review their practices regularly, monitor developments in every place they call, and take legal advice when launching or changing AI calling programs.

Honesty as the Foundation

Across every country and law, the same themes appear: people should agree to automated calls before receiving them, and they should know when they're talking to a machine. Businesses that disclose AI clearly, obtain proper consent, avoid impersonation, and keep people in control are not only far better placed to comply with today's laws, but also with the rules still being written. As AI voices become harder to distinguish from human ones, honesty about who, or what, is on the line becomes more important, not less.

Related articles

• The TCPA Explained: The Main US Law on Phone Calls and Texts
• The Telemarketing Sales Rule: What It Requires and Who It Covers
• Do Not Call Registries: How They Work in the US and UK