FAQ privacy
AI Regulation (AI Act)
Which provisions of the EU AI Act apply to AI Agents and what risk category do they fall into?
sipgate AI Agents generally fall under the category of "limited risk" pursuant to Art. 50 of the AI Regulation. For systems in this category, transparency obligations primarily apply. The AI Regulation classifies systems for direct interaction with natural persons (such as Voice Agents) as subject to transparency obligations, but not as high-risk systems. The AI Act has been in force since 1 August 2024; the transparency obligations under Art. 50 apply from 2 August 2026.
Must the AI Agent actively point out at the beginning of a conversation that it is an AI-supported dialogue, and in what form?
Yes, under Art. 50(1) of the AI Act, providers of AI systems intended for direct interaction with natural persons must design them so that the persons are informed that they are interacting with an AI system. The notice must be clear, unambiguous, and transparent, at the latest upon the first interaction. An exception applies only if this is obvious from the context. sipgate ensures by default that the AI Agent outputs a pre-formulated notice (greeting text) at the beginning of each conversation that clearly discloses the AI-supported nature of the dialogue. This notice is technically fixed at the start of the conversation and cannot be skipped. The notice must be reproduced in the specified wording regardless of the language setting and meets accessibility requirements.
What fines apply for violations of the transparency obligations of the AI Act?
Violations of the transparency obligations under Art. 50 of the AI Act can, pursuant to Art. 99(4) of the AI Act, be punished with fines of up to EUR 15 million or 3% of the company's worldwide annual turnover (whichever is higher). Fines are imposed by national supervisory authorities. It should be noted that responsibility for compliance with the transparency obligations lies with the operator of the AI system (the customer), while sipgate, as the provider of the AI system, supplies the technical requirements.
Data protection (GDPR)
What legal basis applies to the processing of conversation data by the AI Agent?
The processing of conversation data may be based, for example, on Art. 6(1)(a) GDPR (consent) or Art. 6(1)(b) GDPR (performance of a contract), depending on the specific use case. For inbound calls in customer service, processing may be necessary for the performance of a contract or for taking steps prior to entering into a contract.
Is the caller's voice biometric data within the meaning of Art. 9 GDPR, and does a company need explicit consent for this?
The human voice can be classified as biometric data within the meaning of Art. 4 No. 14 and Art. 9 GDPR if it is processed using special technical procedures for the unique identification of a person. The purpose of the processing is decisive: if the voice is used solely for speech recognition and conversation handling (without biometric identification), Art. 9 GDPR does not apply. sipgate AI Agents use voice data exclusively for conversation processing, not for biometric identification. Explicit consent pursuant to Art. 9(2)(a) GDPR is therefore not mandatory.
Data processing
Does sipgate use our customers' data for training or machine learning?
No. sipgate does not use customer data for the training of AI models or machine learning. This is contractually excluded with all AI service providers (especially OpenAI). The data is used exclusively to provide the commissioned services and is deleted or anonymized after processing is complete. This regulation is part of the technical and organizational measures and documented in the DPA.
Can sipgate indemnify us against liability to third parties, especially in the event of violations of the AI Act and GDPR?
An indemnity by sipgate is not provided for, since the customer, as controller, is generally responsible for the lawful use of the AI Agent. This corresponds to the statutory allocation of roles under the GDPR and AI Act. sipgate undertakes, however, to provide the technical and organizational prerequisites for lawful use and to support the customer in complying with its obligations under the provisions of the DPA. The systems provided are generally designed so that they can be used in compliance with the GDPR: (1) In our data processing agreement (DPA), the details of data processing are regulated. Appropriate agreements exist with our subprocessors and partners to ensure lawful data processing. (2) We also point out that the assistant clearly points out or can point out at the beginning of each conversation that it is AI (Further information can be found in our privacy policy as well as in the data processing agreement) In cases where a violation is due to errors or breaches of duty by sipgate, sipgate shall be liable in accordance with the contractual liability provisions.
Liability & Responsibility
Who is liable if the AI Agent makes mistakes, gives false information, or violates the AI Regulation or GDPR?
Liability is determined by the division of roles between controller and processor: The customer, as the operator of the AI Agent, is the controller within the meaning of the GDPR and the AI Act and is therefore generally responsible for lawful use. This includes, in particular, compliance with transparency obligations, lawful data processing, and ensuring that the AI Agent is configured appropriately for its intended use. sipgate is liable as a processor for breaches of duty in the context of data processing (Art. 82 GDPR).
Last updated