Yes, an AI phone assistant can be GDPR-compliant. What matters is not the AI itself but how you use it: callers need to know they are talking to an AI and what happens to their data, you need a legal basis and a data processing agreement, and the data should stay in the EU.
This article gives you a to-do list instead of a collection of legal paragraphs: what you have to handle yourself, what your provider must deliver and a sample greeting for your phone line. A quick note up front: this is not legal advice. For your specific case, please talk to your data protection officer.
The article is written with German businesses in mind, so some of the laws mentioned (such as the German Criminal Code) apply in Germany specifically.
Short answer: when is an AI phone assistant GDPR-compliant?
You can use an AI phone assistant in a privacy-compliant way when these five points are in place:
- Transparency: The AI identifies itself as an AI at the start of the call and points to the privacy notice.
- Legal basis: For every processing step (call, transcript, summary) it is clear which GDPR article it relies on.
- Recording only with consent: If the call is recorded, you get consent first.
- Contract with the provider: There is a data processing agreement (DPA) under Art. 28 GDPR.
- Data location and security: The data is processed in the EU, encrypted and not kept longer than necessary.
What data does an AI phone assistant actually process?
Before getting into legal articles, it helps to look at what a call actually produces. Typically:
- Phone number and connection data: who called, when and for how long.
- The voice: the spoken word the AI understands in real time.
- Call content: name, request, customer number, preferred appointment, sometimes sensitive details too.
- Transcript and summary: the written version of the call and a short version for your team.
- Context from your systems: for example a calendar entry or a ticket created after the call.
All of this is personal data. That is why the principles of Art. 5 GDPR apply: collect only what is needed, use it only for the stated purpose, keep it no longer than necessary and protect it appropriately.
The voice itself is a special case. If it is used to uniquely recognise people (that is, as a biometric feature), the stricter Art. 9 GDPR applies. According to Dr. Datenschutz (in German), you will usually need explicit consent for that. If an AI recognises returning callers by their phone number and your customer data instead, that is not voice biometrics. Ask your provider which approach is used.
On what legal basis may the AI handle calls?
Every data processing step needs a legal basis under Art. 6 GDPR. For an AI phone assistant, three are especially relevant:
- Contract or pre-contractual steps (Art. 6(1)(b)): Someone calls to book an appointment or sort out an order. The data is needed to do exactly that.
- Legitimate interest (Art. 6(1)(f)): You want to be reachable and document requests properly. That can work for pure service processes, but it requires a careful balancing of interests, as Dr. Datenschutz stresses.
- Consent (Art. 6(1)(a)): The safest basis when the call is recorded and analysed by the AI. The law firm IT-Recht Kanzlei (in German) recommends explicit consent from callers for AI telephony.
Whatever the basis, Art. 13 GDPR requires you to inform callers: who is responsible, what the data is used for, how long it is stored and what rights callers have. On the phone this works best in two steps: a short notice in the greeting and the full details on a web page.
Important for outbound calls: if you use the AI to call people proactively to sell something, German law (§ 7 UWG) generally requires prior explicit consent from consumers.
May the AI record and transcribe calls?
Yes, but not secretly. Under the German Criminal Code, § 201 StGB makes it a criminal offence to record another person’s privately spoken words without authorisation. The penalty can be up to three years in prison. So anyone who records needs the caller’s consent, and before the recording starts.
Pure live transcription without a stored audio recording is more of a grey area. In a reply on its site, Dr. Datenschutz refers to an assessment by the Baden-Württemberg data protection authority, according to which at least implied or presumed consent may be needed here too. The pragmatic approach: treat transcription like recording and inform callers in advance.
What could the greeting sound like?
IT-Recht Kanzlei recommends this order before the AI gets going: welcome, notice about the AI and recording, reference to a page with the privacy details, then consent. If you also offer a human on the phone, callers can choose by pressing a key. If there is only the AI, it is enough to say that continuing the call counts as consent.
This is how it could sound (sample text, please check it with your data protection officer):
“Hello and welcome to Sample Ltd! I’m the team’s digital assistant, which means I’m an AI. To handle your request, this call will be recorded and analysed. You can find all the details at sample.com/phone-privacy. If that’s fine with you, just stay on the line. If you’d rather speak to a person, press 0.”
Want a more formal or more casual tone? Simply adapt the greeting to the way you talk to your customers.
Do I have to tell callers they are talking to an AI?
Yes. Since 2 August 2026, the transparency obligation under Art. 50 of the AI Act applies. AI systems that interact directly with people must be designed so that people learn they are talking to an AI. The exception: it is already obvious to a reasonable person from the circumstances. With a natural-sounding voice on the phone, that is rarely the case.
Under Art. 50(5), the notice must be clear and distinguishable “at the latest at the time of the first interaction”. On the phone that means: in the first sentences of the greeting, not somewhere in the privacy policy.
Didn’t the Digital Omnibus postpone this?
No, not for this obligation. With the Digital Omnibus on AI (Regulation (EU) 2026/1744) of 8 July 2026, the EU mainly postponed the obligations for high-risk AI. The Art. 50 transparency obligations for normal operation still apply from 2 August 2026, as nachweisfertig.de and the Regulatory Blog (both in German) summarise. A transition period until 2 December 2026 only covers the machine-readable labelling of AI-generated content (Art. 50(2)) for systems already on the market before that date.
Many guides still talk about it in the future tense, “from August 2026”. For you, the obligation is already here.
What does my provider need to deliver: DPA, server location, security?
Your provider processes call data on your behalf. That requires a data processing agreement (DPA) under Art. 28 GDPR. IT-Recht Kanzlei simply calls it “unavoidable” for AI telephony. Among other things, the DPA lists which sub-processors (for example for language models or telephony) are involved.
Server location and third-country transfers: Data may only leave the EU under the conditions of Art. 44 ff. GDPR. That is possible, but it makes things more complicated. Depending on the provider, individual components, such as the language model, may run outside the EU. So don’t just ask where the provider is based; ask where speech processing, the language model and storage actually run.
Technical security: Art. 32 GDPR requires appropriate safeguards, such as encryption, access controls and two-factor sign-in.
Training on your data: Find out whether calls are used to train AI models. According to IT-Recht Kanzlei, such data must at least be pseudonymised first.
What you handle and what the provider delivers
| Task | You as the business | Your provider |
|---|---|---|
| Greeting with AI notice and consent | write and approve it | implements it technically |
| Privacy information (Art. 13) | create the web page | supplies the processing details |
| DPA (Art. 28) | review and conclude it | provides it |
| Retention period | decide on it | makes it configurable |
| Data protection impact assessment (Art. 35) | check whether it’s needed and carry it out | supplies information on technology and risks |
| Security (Art. 32) | assign access within your team | ensures encryption, EU hosting, MFA |
On the impact assessment under Art. 35 GDPR: Dr. Datenschutz considers it “almost always necessary” in practice for AI in customer service.
What applies to medical practices, tax advisers and law firms?
Anyone bound by professional secrecy needs to look even more closely. In Germany, § 203 StGB protects secrets entrusted to, for example, doctors, lawyers or tax advisers. Two points matter for you:
- Bringing in service providers is allowed, but only as far as necessary for their work (§ 203(3)).
- You must bind them to confidentiality. If you don’t and the provider discloses something, you can be liable yourself (§ 203(4)).
On top of that comes the GDPR: if a patient mentions symptoms or a diagnosis on the phone, that is health data under Art. 9 GDPR, which is subject to stricter rules. In practice: let the AI capture only the organisational details (name, callback number, preferred appointment) and don’t have it actively ask for medical details. Also make sure summaries only reach the people who need them.
Good to know: “GDPR-compliant” is not a seal a tool simply has. The provider can deliver privacy-friendly technology, but you remain responsible for how it is used. Only the combination of good technology plus your greeting, your privacy information and your DPA makes an AI phone assistant truly privacy-compliant. The German Data Protection Conference (DSK) has published an orientation guide on AI and data protection (in German), intended as a checklist for businesses.
How does qutee.ai handle this?
qutee.ai is an AI phone assistant from Germany, built with these requirements in mind. Specifically (details under Security & data controls and in the Trust Center):
- Data stays in the EU: Processing takes place exclusively in the EU or EEA. Moving it to a third country is only possible with your prior consent. You can find every service provider in the list of sub-processors.
- Greeting on your terms: You adjust the voice, greeting and behaviour, including the AI notice and the reference to your privacy information.
- A human when needed: Calls can be forwarded to the right person, and the assistant can behave differently depending on your business hours.
- Data control: You manage access, retention and organisation security yourself, including a configurable retention period and two-factor sign-in.
- DPA included: You conclude the data processing agreement under Art. 28 GDPR electronically when you sign up, without a separate signature.
Want to hear what a transparent AI assistant on the phone feels like? Call one of our live demos or try qutee.ai free for 14 days. If you have questions about your data protection setup, simply book an intro call with us.
FAQ: AI phone assistants and GDPR
It can be. You need a legal basis, an AI notice at the start of the call, privacy information under Art. 13 GDPR, a DPA with the provider and secure processing, ideally in the EU.
Yes. Under Art. 50 of the AI Act, people must have been told since 2 August 2026 that they are interacting with an AI, at the latest at the first interaction. On the phone, the notice therefore belongs in the greeting.
Only with the caller's consent. In Germany, secret recordings are a criminal offence under § 201 StGB. A common approach is a notice in the greeting stating that continuing the call counts as consent, or a choice by pressing a key.
Yes. The provider processes call data on your behalf, so a DPA under Art. 28 GDPR is mandatory. Among other things, it covers security, sub-processors and deletion.
Yes, with extra care. In Germany, the provider must be bound to confidentiality under § 203 StGB, and health data falls under Art. 9 GDPR. Ideally the AI captures only organisational details.
Often, yes. For AI in customer service, data protection experts consider an impact assessment under Art. 35 GDPR almost always necessary in practice. Check this with your data protection officer.
As of November 2026. This article is not legal advice.



