Why the EU AI Act is relevant for Contact Centers now

The EU AI Act (Regulation (EU) 2024/1689, "EU AI Act") is the first comprehensive legal framework for artificial intelligence. It applies step-by-step — and a central building block directly affects contact centers: the transparency obligations under Article 50. Anyone who uses voicebots, chatbots, AI-generated content, or emotion recognition must disclose that AI is involved.

For operators of customer service and contact center environments, this means: It is no longer just about service quality, but about a documented legal obligation with a risk of fines.

The new timeline after the Digital Omnibus

In July 2026, the EU postponed parts of the original deadlines with the Digital Omnibus (Regulation (EU) 2026/1744, in force since July 27, 2026) — in particular, the obligations for high-risk systems. Important: The transparency obligations under Art. 50 were not postponed.

Since February 2, 2025, prohibited practices under Art. 5 have applied, including emotion recognition in the workplace. Since August 2, 2025, obligations for general purpose AI models (GPAI) have applied. Since August 2, 2026, transparency obligations under Art. 50 for chatbots, deepfakes, emotion recognition, and AI content have applied. From December 2, 2026, the labeling and watermarking obligation for AI-generated content under Art. 50 Paragraph 2 takes effect. The high-risk obligations under Annex III were postponed to December 2, 2027 by the Digital Omnibus, and AI systems in regulated products under Annex I to August 2, 2028.

In short: The relief provided by the Digital Omnibus affects the complex high-risk obligations. The disclosure obligations in customer contact already apply now.

Article 50: Four transparency obligations in customer contact

Art. 50 obligates providers and operators regardless of whether the system is classified as high-risk. Four areas are relevant for contact centers.

Direct interaction: Labeling chatbots and voicebots

When customers speak or chat with an AI system, they must be informed that they are interacting with an AI — unless this is obvious from the context. For voicebots, this means a clear announcement or disclaimer at the start of the call.

Labeling AI-generated content

Synthetically generated text, audio, image, or video content must be labeled as artificially generated in a machine-readable format. This watermarking obligation under Art. 50 Paragraph 2 takes effect from December 2, 2026.

Emotion recognition: Particular caution

If a system is used for emotion recognition, the affected persons must be informed. Warning: Emotion recognition is generally prohibited in the workplace and in educational institutions (Art. 5) — with narrow exceptions for safety or medical reasons. For contact centers, this is particularly delicate if AI is intended to analyze the mood of their own agents.

Deepfakes and AI texts on public issues

AI-generated deepfake content and published AI texts on matters of public interest must be labeled as such.

What this means in practice for contact centers

The practical core for customer service operators: Voicebots and chatbots must be disclosed at the start of every AI interaction, so it is clear that no real person is responding. The separation of customer and agent emotions is important — the AI-supported sentiment analysis of customer conversations is possible with information and a legal basis, whereas the analysis of employee emotions in the workplace is regularly inadmissible. Automatically generated responses or summaries must be identified as AI-generated, and all information processes should be comprehensibly documented.

Fines: Why this is not a marginal topic

Violations of the transparency obligations under Art. 50 can be punished with fines of up to 15 million Euros or 3% of global annual turnover, whichever is higher. For prohibited practices under Art. 5, the framework is even up to 35 million Euros or 7% (Art. 99). This is enforced by national market surveillance authorities.

Implementation in practice

A clear process is recommended for implementation: First, create an AI inventory — where are voicebots, chatbots, AI summaries, and sentiment analyses running? Next, integrate disclosure, i.e., an audible or visible notice that AI is being used. Then, review emotion recognition: If employees are analyzed, the application must be stopped or legally clarified. Additionally, prepare the labeling of AI-generated content (watermarking from December 2026), document all processes, times, and responsibilities, and regularly monitor AI outputs for accuracy and non-discrimination.

How onsoft supports you with AI transparency

onsoft combines call recording, AI speech analysis, and quality management in an audit-proof system. This allows AI-supported evaluations in the contact center to be comprehensibly documented — including proof that customers were informed about the use of AI. For voicebots, continuous quality and transparency control helps to consistently ensure disclosure and response quality.

Frequently Asked Questions

Do the transparency obligations of the EU AI Act already apply despite the Digital Omnibus?

Yes. The Digital Omnibus primarily postponed the high-risk obligations. The transparency obligations under Art. 50 have applied unchanged since August 2, 2026.

Do I have to disclose that a voicebot is not human?

Yes. Users must be informed that they are interacting with an AI system, unless this is obvious. For voicebots, this typically occurs via an announcement at the start of the call.

Is emotion recognition allowed in the contact center?

In the workplace towards one's own employees, emotion recognition is generally prohibited (Art. 5), with narrow exceptions. The analysis of customer emotions is possible with information and a legal basis.

How high are the fines under the EU AI Act?

For violations of Art. 50, up to 15 million Euros or 3% of global annual turnover; for prohibited practices, up to 35 million Euros or 7%.

From when must AI content be labeled with a watermark?

The machine-readable labeling of AI-generated content under Art. 50 Paragraph 2 takes effect from December 2, 2026.