Before an AI agent texts or calls your customers: the consent rules in Australia, the US, the UK and the EU
Consent, opt-out, sender registration and AI disclosure for AI agents that text or call customers in Australia, the US, the UK and the EU. Sourced, not legal advice.
Ignacio Puig Moreno
Co-founder · Business
Before an AI agent texts or calls a customer, check four things: consent to contact them, an opt-out that works, sender registration where it exists, and a clear statement that it is an AI. Australia, the US, the UK and the EU differ on the details but agree on the first two.
This post summarises public sources and is not legal advice. I wrote it because the first question after "can an agent text our customers?" is always "is that allowed?", and the honest answer is that it depends on the country, the channel and what the message is for.
What do the four places have in common?
Consent comes first, and the person has to be able to say stop. Everything else varies. The table lists what we could verify at the source for each one, and leaves a cell empty when we did not verify it.
| Consent to contact | Opt-out | Sender identity | AI disclosure | |
|---|---|---|---|---|
| Australia | Required for each person, with evidence if the regulator asks | Actioned within 5 working days, free, works for 30 days after sending, no login | Branded sender IDs must be registered from 1 July 2026 | Not covered here |
| United States | Prior express consent for artificial-voice calls, which includes AI voices | Any reasonable means, and replying STOP always counts | Not covered here | Not covered here |
| United Kingdom | Required for individuals, with a soft opt-in for your own customers | Not covered here | Not covered here | Not covered here |
| European Union | Prior consent for automated calls and electronic mail used for direct marketing | A valid address to ask that messages stop | Sender identity cannot be hidden | Required from 2 August 2026 |
What does Australia require?
The ACMA says you must have the consent of each person you message and be able to produce the evidence if asked. Commercial messages must carry an unsubscribe option that is actioned within 5 working days, costs nothing, works for at least 30 days after the message was sent and does not require the person to log in or provide more personal information.
There is also a sender rule. From 1 July 2026, texts that carry an organisation's name at the top need that sender ID registered with the ACMA.
What does the United States require?
Two rulings matter for an agent. First, the FCC decided in February 2024 that an AI-generated voice is an artificial voice under the Telephone Consumer Protection Act. The Chairwoman put it this way:
calls that use this technology to simulate a human voice are illegal, unless callers have obtained prior express consent.
The ruling itself is narrower than that headline. As a law firm's reading points out, AI voice calls are not banned: they fall under the same consent, identification and opt-out rules as other artificial or prerecorded voice calls.
Second, the FCC's revocation rule says people can withdraw consent by any reasonable means, and a business cannot name one exclusive method. A reply of "stop", "quit", "end", "revoke", "opt out", "cancel" or "unsubscribe" to a text is a reasonable means in itself. The rule took effect on 11 April 2025, and a revocation applies to both robocalls and robotexts whichever medium it came through.
What do the United Kingdom and the EU require?
In the UK, the ICO explains that marketing texts and emails to individuals need consent. A soft opt-in lets you contact your own customers about similar products and services, and it does not apply to prospects or to bought lists. The consent rule does not apply to corporate subscribers.
In the EU, Article 13 of the ePrivacy Directive allows automated calling systems and electronic mail for direct marketing only with the subscriber's prior consent. The Directive's recital 40 names SMS, and its definition of electronic mail covers text and voice messages. It also prohibits hiding the sender's identity and requires a valid address for asking that messages stop.
On top of that, Article 50 of the AI Act requires AI systems that interact directly with people to tell them so, unless it is obvious, from 2 August 2026.
What does this change for an AI agent in practice?
Four design decisions follow, and none of them is legal advice.
- Record the consent per person and per channel. Australia expects you to produce it.
- Make the agent recognise an opt-out in any wording, not only the word STOP. The US rule counts any reasonable means.
- Carry the opt-out across channels. In the US a revocation covers both calls and texts.
- Say it is an AI at the start, by text and by voice. We cover the timing in a tool on when the agent tells people.
WhatsApp adds a platform layer on top of the law. Meta requires opt-in and limits free replies to a 24-hour window.
Who signs off?
Your legal and compliance team, in each country you operate in. Whether a message counts as marketing or as a service notice, for example, is a question for counsel. What we can do is build the agent so the answers you decide on are the ones it follows, and so you can show a record of what it did.
Frequently asked questions
Do AI voice calls need consent in the United States?
Yes. In February 2024 the FCC ruled that AI-generated voices are an artificial voice under the Telephone Consumer Protection Act, so the Act's consent, identification and opt-out rules apply to them. The ruling does not ban such calls: it places them under the same rules as other artificial or prerecorded voice calls.
Must an AI agent tell people it is an AI in the EU?
Article 50(1) of the EU AI Act requires systems that interact directly with people to be designed so that people are informed they are dealing with an AI, unless that is obvious from the circumstances. The provision applies from 2 August 2026. The Commission's guidance says the notice must come at the start of the first interaction, clearly and distinguishably.
Does StudioChat make my agent compliant?
No tool can guarantee that. We give you guardrails, escalation rules and a record of every conversation, and the decision about what is compliant in your country stays with your legal and compliance team. This post summarises public sources and is not legal advice.
Ignacio Puig Moreno · Co-founder · Business, StudioChat
Want agents like these working for your team?
Talk to us now