Guide · AI Phone Calls

Called by an AI? What's behind AI phone calls, and what's allowed.

An AI phone call is a conversation where an artificial intelligence speaks instead of a human. Behind it can be something useful, such as an appointment confirmation or a business calling you back, or a scam attempt using a cloned voice. This page sorts out both worlds: what's allowed, what a legitimate call looks like, and when to hang up.

By — AI Automation Manager (TÜV SÜD Akademie), three IHK qualifications, trained HVAC and plumbing tradesman and field service technician.

Published

Why the topic feels so muddled right now

Search for AI phone calls and you'll find two kinds of results that have little to do with each other: warnings about phone scams using artificial voices on one side, and marketing pages for phone software on the other. Both are true, and both are only half the story. The same technology that can recreate a voice convincingly is, in many businesses, simply answering calls when no one is free to pick up.

It's worth keeping the two worlds separate. Whether an AI is speaking on the phone says little by itself. What matters more is who's calling and what they want from you.

Two worlds: legitimate AI calls and scam tactics

The differences are clearer in everyday life than you might think. A side-by-side of the typical features:

Typical features compared. Not a statistic, but patterns from practice.
Feature Legitimate AI Call Scam Call
Who is calling A business you know or contacted yourself An unknown or withheld number, claiming to be a bank or authority
How the voice introduces itself As an AI, with a company name and reason for calling As a human, often posing as a relative or an official
What it's about An appointment, a callback, an order: traceable and concrete Money, one-time codes, passwords, gift cards
How much pressure builds up None. You can hang up and call back later Time pressure and fear: act now, tell no one
What a callback reveals The number clearly belongs to the business Calling the official number back exposes the call

The most important feature is in the second row: a properly set-up AI introduces itself as an AI at the start of the call, a scammer does the opposite. The check steps for a live call, from voice to follow-up question, are in the checklist Detecting AI Calls.

What's allowed: the transparency requirement in the EU AI Act

AI on the phone isn't banned in Germany. But the EU AI Act requires that people be able to recognize when they're interacting with an AI system. In practice: the AI identifies itself as an AI at the start of the call instead of pretending to be human. On top of that come the usual rules that apply with or without AI: personal data must be processed in compliance with data protection law, and unsolicited telemarketing stays tightly restricted.

For what data protection looks like for an AI phone assistant in practice, from data processing agreements to call recording, see the article Is an AI phone assistant GDPR compliant?

The legitimate side: what businesses use AI calls for

The most common use is unspectacular: an AI phone assistant answers incoming calls when no one in the business is free to pick up. It answers recurring questions, logs the caller's request along with a callback number, and hands everything over to the team, sorted. There are also appointment confirmations and announced callbacks, for example when you've asked for a callback on a website. Our assistant Steffi is built for exactly these tasks: she captures what the caller needs.

What all of these uses have in common: the caller already knows the business, or reached out themselves, and the AI doesn't hide what it is. One example from practice: the phone keeps ringing at a law firm even during a client meeting, and here's how the AI call handling for law firms catches it. What a call like that sounds and feels like is described in the article Talking to AI on the phone. The technology behind it, from speech recognition to speech synthesis, is explained in the foundational article on voice AI, and how a voice agent works technically, in detail, in the article on the voicebot.

The scam side: cloned voices, old tricks

The schemes themselves aren't new: the shock call, the grandparent scam, the fake bank employee. What's new is that the voice on the other end can sound convincingly like someone you trust. Cloning a voice today takes very little audio, for example from voice messages or videos on social media.

Luckily, your defense doesn't depend on hearing that a voice is artificial. It lies in your behavior: hang up and call back yourself, at the number you already have. Never give out one-time codes, passwords or PINs on the phone, no matter how urgent it sounds. And agree on a codeword within your family that you ask for with unusual requests. A real relative knows it, a voice clone doesn't.

Note: this is not legal advice

This article explains the topic in general terms and doesn't replace legal advice. Whether a specific use is permitted depends on the individual case. For a binding assessment, contact a data protection officer or a specialized lawyer. If you've suffered damage from a scam attempt, the police can help.

Frequently asked questions

The most common questions about AI phone calls.

What are AI phone calls?

AI phone calls are phone conversations where an artificial intelligence speaks instead of a human. That can be a legitimate call, such as an appointment confirmation or a business calling you back, but also a scam attempt using an artificially generated or cloned voice. Whether it's a useful service or a scheme depends on who's calling and what they want.

Are AI phone calls legal in Germany?

Yes, in principle. A business may put an AI on the phone as long as it follows the rules: callers should be able to recognize they're speaking with an AI, personal data must be processed in compliance with data protection law, and unsolicited telemarketing stays tightly restricted, with or without AI. What's banned is deception, not the technology.

Does an AI on the phone have to say it's an AI?

The EU AI Act requires that people be able to recognize when they're interacting with an AI system. In practice, that means a properly set-up AI introduces itself as an AI at the start of the call. That's exactly what separates legitimate calls from illegitimate ones: scammers do the opposite and pass their voice off as human.

What is voice-cloning fraud?

In voice cloning, a voice is artificially recreated, often from just a few seconds of audio. Scammers use this to pose as relatives, supervisors or bank employees on the phone and demand money or access credentials. Typical signs are time pressure, staged emotional emergencies, and a request not to tell anyone.

How should I react to a suspicious AI call?

Hang up, take a breath, and call back yourself: the person or business at the number you already have, not the number the call came from. Never give out access credentials, one-time codes or PINs over the phone, no matter how plausible the story sounds. For families, an agreed-upon codeword that only they know helps.

What do businesses legitimately use AI calls for?

Mainly for availability: the AI answers incoming calls when no one is free to pick up, answers recurring questions and logs callback requests. Appointment confirmations and announced callbacks belong here too. What these uses have in common: the caller already knows the business, and the AI doesn't hide what it is.

More questions? Find all the answers in our FAQ →