Callbot & IA vocale

Callbots and GDPR: are they compatible?

Callbots and GDPR: informing callers, retention, hosting, processors and the AI Act. The points to check for a compliant voice agent.

Thomas Guenoux
Thomas Guenoux
25/08/2026
·
5
min read
Callbots and GDPR: are they compatible?

A callbot handles calls, and therefore personal data: phone number, voice, identity, contract number, reason for calling. Before deploying a voice agent on your phone line, one question always comes up with DPOs and CISOs: is a callbot compatible with GDPR? Yes, provided a few specific rules are followed, by both the company and the vendor.

The article in brief
A callbot can be GDPR compliant if the caller is informed, if the data collected is limited to what is necessary, if its retention period is defined and if it is properly hosted and protected. The company remains the data controller; the callbot vendor acts as a processor, bound by a data processing agreement (DPA). Since 2 August 2026, the AI Act also requires informing callers that they are talking to an AI.

Why GDPR applies to callbots

As soon as a person can be identified, directly or indirectly, GDPR applies. Over the phone, this is almost always the case: the calling number, name, address, customer number or the voice itself are personal data. A callbot collects this information to understand the request, find the file and respond.

Some data requires particular care: health data (health insurers, healthcare facilities), banking data, or the voice when it is used to authenticate the person (biometrics). Processing them is subject to stricter conditions.

Who is responsible for what?

  • The company deploying the callbot is the data controller. It defines the purposes (answering requests, booking appointments, qualifying calls), chooses the legal basis, informs callers and responds to their requests to exercise their rights.
  • The callbot vendor is a processor within the meaning of Article 28 of GDPR. It processes data on behalf of the company and according to its instructions. Its commitments are set out in a data processing agreement (DPA): security measures, data location, retention periods, sub-processors, assistance in the event of a breach.

The points to check for a compliant callbot

1. Inform the caller

Callers must know that their data is being processed, why, by whom and for how long. Over the phone, this information is usually given in two steps: a short message at the start of the call, then a link to a detailed page (privacy policy). If the call is recorded, the caller must be informed before the end of the conversation, with the option to object.

2. Collect only what is necessary

This is the data minimization principle. The callbot should only ask for information that is useful for handling the request. Clearly defining the agent's scope (what it handles, what it transfers) also limits the data collected.

3. Set retention periods

Transcripts, recordings and technical logs must have a defined and justified retention period. For call recordings, the CNIL (the French data protection authority) has recommended keeping recordings for no more than six months and analysis documents for no more than one year (its guidance on the subject is being updated).

4. Control hosting and transfers

Where are conversations stored? Does data leave the European Union, in particular to speech recognition or language model providers? These points must be covered in the DPA and in the vendor's list of sub-processors.

5. Secure the data

Encrypted exchanges, access control based on the principle of least privilege, strong authentication for sensitive access, logging and backups: the vendor must be able to describe its measures and, ideally, prove them (certifications, audits).

6. Enable the exercise of rights

Callers can request access to their data, its correction or its deletion. The company must be able to find the conversations concerned, and the vendor must help it respond within the deadlines.

7. Govern the use of data for AI

Are your customers' conversations used to train models? On whose behalf? The answer must be clear and contractual. An architecture that relies on your knowledge base, without retraining a model on your data, greatly limits this risk.

GDPR and the AI Act: an additional transparency obligation

The European regulation on AI (AI Act) adds to GDPR. A customer service callbot generally falls into the limited-risk category. The main obligation is transparency: since 2 August 2026, people must be informed that they are interacting with an AI system, unless it is obvious. In practice, the welcome message must state that the caller is speaking to a virtual assistant.

How YeldaAI meets these requirements

YeldaAI voice agents are designed for demanding sectors such as banking, insurance, healthcare and public services:

  • Hosting in France with Clever Cloud (ISO 27001, HDS, SecNumCloud), with no transfer outside the European Union.
  • Defined retention periods: conversations kept for 6 months after the last conversation, technical logs for 3 months, periods adjustable by contract.
  • No model training on your data: YeldaAI does not train its own language models; the agent relies on your knowledge base.
  • A processor within the meaning of Article 28, with a signed DPA for each customer and a DPO reachable at dpo@yelda.ai.
  • A controlled scope: the agent only acts within the framework you define, every conversation is logged and out-of-scope calls are transferred to your advisors.

Details of our measures, the list of our sub-processors and the documents available on request are gathered in our Trust Center.

Conclusion

Callbots and GDPR are compatible, provided compliance is addressed from the design stage: informing callers, minimization, retention periods, hosting, security and transparency about AI. The choice of vendor plays a central role, since it implements a large part of these safeguards. Before signing, ask for its DPA, its list of sub-processors and where the data is located.

Preparing a voice agent project with your DPO? Let's discuss your requirements.

How many calls can your voice agent absorb?

Estimate in one minute the number of automated calls and the savings on your own volumes, with our ROI calculator.

Read also

Our latest articles

All articles