Outbound AI calls: consent, disclosure and calling rules
The moment a voice agent places a call instead of answering one, it is regulated differently. Most of the rules were written for robocalls and telemarketing, and regulators have been clear that an AI voice does not escape them by sounding human. In practice four things decide whether an outbound campaign is lawful: whether the person consented to this kind of call, whether the call identifies itself and discloses that it is automated, when and how often it calls, and how quickly an opt-out takes effect. This guide covers the United States, the United Kingdom and the European Union at the level a business needs to plan a campaign. It is not legal advice, and a campaign of any size should be checked by a lawyer for the places you call.
Does the TCPA apply to AI voice agents?
Yes. In February 2024 the US Federal Communications Commission ruled that calls using AI-generated voices count as calls using an "artificial or prerecorded voice" under the Telephone Consumer Protection Act. That brings AI voice calls under the same consent rules as robocalls, whatever the voice sounds like and however conversational the agent is.
The practical consequence is consent. Calls to mobile numbers and residential lines using an artificial voice generally need the called person's prior express consent, and telemarketing calls, those selling something, need prior express written consent. A customer who gave you their number and agreed to reminders can receive a reminder; a number bought on a list cannot receive a sales pitch.
The TCPA allows people to sue for violations, and damages are set per call, which is why a campaign that is wrong at scale becomes expensive quickly. Several states add their own telemarketing rules on top of the federal ones, some stricter than the TCPA, so the state of the person you are calling matters too.
What must the call say and when can it be made?
Federal rules require an artificial-voice call to identify the business responsible at the start and to give a telephone number for it. Telemarketing calls must also offer a way to opt out during the call itself, which for a voice agent means recognising "stop calling me" or an equivalent and acting on it immediately.
Telemarketing calls may only be placed between 8am and 9pm in the called person's local time, and numbers on the National Do Not Call Registry must not receive telemarketing calls without consent or an existing business relationship that permits it. Your own internal do-not-call list must be honoured as well, and kept for as long as the rules require.
The simplest safe design follows the strictest version everywhere: identify yourself and say the caller is an AI assistant in the first sentence, give a callback number, call only in daytime hours in the recipient's time zone, and treat any request to stop as permanent across every campaign.
| Requirement | United States | United Kingdom | European Union |
|---|---|---|---|
| Consent for automated marketing calls | Prior express written consent (TCPA) | Prior consent for automated calls (PECR) | Prior consent under national ePrivacy rules |
| Service calls such as reminders | Prior express consent for artificial-voice calls | Not marketing, but a lawful basis under UK GDPR | Not marketing, but a lawful basis under GDPR |
| Do-not-call lists | National DNC Registry plus your own list | TPS and CTPS for live marketing calls | Varies by country |
| Say it is automated | Identify the business and give a number | Identify the caller; say it is automated | AI Act: tell people they are talking to an AI |
| Calling hours | Telemarketing 8am to 9pm local time | No fixed hours; avoid unreasonable times | Varies by country |
| Opt-out | In-call opt-out for telemarketing, honoured at once | Honour objections immediately | Honour objections immediately |
What are the rules in the United Kingdom?
The Privacy and Electronic Communications Regulations treat automated marketing calls, calls that play a message or run without a person on the line, as needing the recipient's specific prior consent. An AI agent that makes marketing calls on its own is likely to be treated as an automated calling system, so the safe assumption is that marketing calls by an AI agent need prior consent.
Live marketing calls must be screened against the Telephone Preference Service and its corporate equivalent. Whether an AI agent counts as a live call is exactly the kind of question to put to a lawyer before a campaign, which is another reason to keep AI outbound calling to people who asked to hear from you.
Calls that are not marketing, such as appointment reminders or service messages, fall outside the marketing rules, but you still need a lawful basis under UK GDPR to use the person's number and data, and you must say who is calling. The Information Commissioner's Office enforces both.
What are the rules in the European Union?
Each member state implements the ePrivacy Directive, which requires prior consent for automated calling systems used for direct marketing, and GDPR governs the personal data involved. Details differ by country, including do-not-call registers and permitted hours, so a campaign across several countries needs checking country by country.
The EU AI Act adds a transparency duty that applies from August 2026: systems intended to interact directly with people must be designed so those people are told they are interacting with an AI, unless it is obvious. For a voice agent that means saying so at the start of the call. It is good practice everywhere and a requirement in the EU.
How do you build compliance into the agent?
Store consent with the contact, including what the person agreed to and when, and have the agent check it before dialling. A campaign should not be able to call a number whose consent does not cover that type of call, which is a data problem to solve in the CRM rather than a line in the script.
Put the hard rules in code, not in the prompt. Calling hours by the recipient's time zone, a maximum number of attempts, and a suppression list checked before every dial are rules a language model should never be trusted to remember. The prompt handles the conversation; the dialler enforces the law.
Record the evidence. Keep the consent record, the call outcome and any opt-out, with timestamps, so that if a complaint arrives you can show why the call was made and what happened. And if you record calls, handle recording consent too, which is covered in its own guide.
Common questions
- Are AI voice calls legal under the TCPA?
- They are legal with the right consent. In February 2024 the FCC ruled that AI-generated voices count as artificial voices under the TCPA, so calls using them to mobile or residential numbers generally need the recipient's prior express consent, and telemarketing calls need prior express written consent. Calls to people who agreed to hear from you, such as reminders they signed up for, can be made lawfully.
- Does an AI calling agent have to say it is an AI?
- It should, at the start of the call. US rules require artificial-voice calls to identify the business and give a contact number, the EU AI Act requires people to be told they are interacting with an AI from August 2026, and UK rules require the caller to be identified. Saying it plainly is also what keeps the listener on the line.
- What hours can outbound AI calls be made?
- In the US, telemarketing calls may only be made between 8am and 9pm in the called person's local time. The UK and EU set expectations of reasonable hours and some countries set specific ones. The safe design is daytime calls only, enforced by the dialler using the recipient's time zone rather than left to the agent.
- Do appointment reminder calls need consent?
- Reminders are not marketing, which lightens the rules, but in the US an AI-voice reminder to a mobile number still generally needs the person's prior express consent, typically given when they booked and shared their number. In the UK and EU you need a lawful basis under GDPR to use their number. Collecting clear consent at booking covers most cases.
- How should opt-outs work for an AI calling agent?
- The agent should recognise a request to stop in natural language, confirm it, end the call and write the opt-out to the contact record so no future campaign, automated or human, calls that number. Telemarketing calls in the US must offer an in-call opt-out. Enforce suppression in the dialler before every call rather than relying on the agent to remember.
