Do You Have to Tell Callers They're Talking to AI? Australian Rules Explained

If you put an AI agent on your phone line, two separate legal questions arise: do you have to say it is AI, and can you record the call? They have different answers.

This is general information, not legal advice. Rules differ by state and are actively changing. Get advice for your situation.

Do you have to disclose that it is AI?

As of now, there is no single Australian law that says you must — but that is a thinner protection than it sounds, for three reasons.

Australian Consumer Law still applies. Misleading or deceptive conduct under s18 does not require a specific AI rule. If your agent actively implies it is human, gives a human name, and a customer relies on that, you are in familiar legal territory using unfamiliar technology.

The direction of travel is clear. The EU AI Act requires disclosure where a person is interacting with an AI system. Several US states have enacted similar rules. Australian reform discussion has consistently included transparency obligations. Building disclosure in now is cheaper than retrofitting it.

Privacy law applies to what the agent collects, regardless of whether disclosure is required — see below.

What good practice looks like

You do not need a legal preamble. What matters is:

• If asked directly, answer honestly. "Am I talking to a person?" gets a plain no. Every agent we build does this.
• Do not construct a false impression. Human names, fake background noise, and scripted hesitation are the features to avoid.
• Make escalation available. A caller who asks for a person should get one.

The commercial argument matters more than the legal one. Callers who discover mid-conversation that they were misled do not become customers. Callers told upfront, who then get their problem solved, largely do not care.

Can you record the call?

This one has actual statutory teeth, and it varies by state.

Australia has federal and state legislation covering surveillance and listening devices. Requirements differ across NSW, Victoria, Queensland, WA, SA, Tasmania, NT and ACT — some are one-party consent, others require all parties.

Practical position: if you record, tell people at the start of the call. It satisfies the stricter states and is the least surprising thing to do.

If the call is inbound from another state, the strictest applicable rule is the safe assumption.

The alternative: do not record

Worth considering, because most businesses record by default without asking whether they need to.

Modern voice platforms offer zero-retention configurations where audio is processed live and nothing is written to storage. No recording, no transcript, no call log.

We run our own website demo this way. For a production agent it is usually the wrong choice — reviewing real calls is how an agent improves, and clients often want the history — but it should be a decision rather than a default.

Privacy obligations either way

Even without recording, an agent that collects a name, number or address is collecting personal information, and the Australian Privacy Principles apply.

• APP 5: tell people what you are collecting and why, at or before collection
• APP 8: if the data goes overseas — and most AI voice processing does — you remain accountable for it
• Retention: decide how long you keep it and delete on schedule

A workable checklist

• Agent admits it is AI when asked, without deflecting
• No human name, no fabricated background noise
• Callers can reach a person on request
• If recording, callers are told at the start
• Retention period decided and enforced
• Privacy policy names what is collected and for how long
• If data goes offshore, that is disclosed

Most of this is a morning's work at setup and effectively free thereafter. Retrofitting it after a complaint is neither.