Europe & regulation
EU AI Act
AI disclosure in dialogue, human oversight, logging and model approval: Univerx is designed for the duties of the AI Act. Which risk class your use case falls into, we work out together.
At a glance
- Designed for
The platform is built toward the requirement. Evidence such as an attestation does not exist yet – we name it as soon as it does.
- On request
The documents exist. You receive them for your review through the contact form.
- Available
Met and evidenced. We assign this level only once the evidence exists.
No documents are listed for this topic. We answer the questions from your questionnaire in conversation.
#What the regulation covers
Regulation (EU) 2024/1689, the AI Act, classifies AI systems by risk. Purpose decides, not technology: the same speech service can be harmless in customer service and sensitive when it shapes a benefits decision.
For customer service this means almost every AI use touches the transparency duties. Some uses can also count as high-risk. That is why it pays to know the classes before a project starts.
#The risk classes
| Class | What it is about | Example in customer contact |
|---|---|---|
| Prohibited practices (Art. 5) | Manipulation, exploiting vulnerabilities, social scoring, emotion recognition at work | Systems that read employees' emotions from biometric data |
| High-risk (Art. 6, Annex III) | Uses with significant consequences, such as access to public benefits, creditworthiness, pricing in life and health insurance | AI that helps decide a benefit claim |
| Transparency duties (Art. 50) | People learn they are interacting with AI; synthetic content is labelled | An AI agent in chat or on the phone |
| Minimal risk | No specific duties under the regulation | Topic detection, summaries for agents |
#How Univerx reflects the duties
Univerx is designed for the duties that apply most often in customer service. The controls are part of the product, not a setting you retrofit later.
#Human oversight in practice
Oversight does not mean reading every sentence the AI produces. It means a person understands what the system does, can step in and can override a decision.
Univerx Assist suggests, the agent decides. Semantic² reports findings, the case officer accepts or rejects them. In both cases the decision stays with a person and is logged.
#Timeline
The duties apply in stages. The prohibitions in Art. 5 have applied since 2 February 2025, the transparency duties in Art. 50 since 2 August 2026.
Amending Regulation (EU) 2026/1744 moved the duties for stand-alone high-risk systems under Annex III to 2 December 2027. The consolidated text on EUR-Lex is authoritative.
How to disclose AI without disrupting the conversation is the subject of Disclosing AI without breaking the conversation. The transparency duties, explained briefly gives the overview. The principles behind it all are on Responsible AI.
#Sources
Products
Where the topic applies.
The product families this topic matters most for.
Univerx AI
AI that resolves requests and backs up people.
- AI Act
- DSGVO
- EU
Univerx Contact Center
Every contact in the right queue.
- DSGVO
- BFSG
- EU
Univerx Trust Center
Security and regulation as a design principle, not an appendix.
- DSGVO
- EU
- AI Act
- BSI C5
- eIDAS
- NIS2
- DORA
- BFSG
Trust & Europe
Every topic in the trust center.
Every topic carries a status. It tells you what is a design goal, which documents you can request and what is evidenced today.
| Topic | Status |
|---|---|
| GDPR | On request |
| Data residency & hosting in Germany | Designed for · German hosting planned |
| EU AI Act | Designed for |
| BSI C5 | Designed for · goal, no attestation |
| eIDAS & online ID | Designed for |
| NIS2 | Designed for |
| DORA | Designed for |
| European Accessibility Act | Designed for · WCAG 2.1 AA as a goal |
| Responsible AI | Designed for · principles & controls |
| Security & availability | Designed for |
| United States: HIPAA, PCI DSS, SOC 2, TCPA, CCPA | Designed for |
Frequently asked questions
Is an AI agent in customer service covered by the AI Act?
Yes, at least by the transparency duties in Art. 50: people must be told they are talking to an AI system unless that is obvious. Whether it is also a high-risk system depends on its purpose, not on the technology.
Who is the provider, who is the deployer?
The regulation distinguishes providers, who place an AI system on the market, from deployers, who use it. If you run AI agents in your customer service, you are a deployer with duties of your own. The roles for your case are set out in the contract.
When do the duties apply?
In stages. The prohibitions have applied since 2 February 2025, the transparency duties in Art. 50 since 2 August 2026. The duties for stand-alone high-risk systems under Annex III were moved to 2 December 2027 in 2026.
Is Semantic² a high-risk system?
It depends on the use. If a public authority uses it to assess entitlement to public benefits, Annex III may apply. Semantic² is designed for human oversight, logging and explainable findings; we work out the classification with you.
Related
Trust & EuropeResponsible AI
Transparency, human oversight, fairness, privacy and governance: the principles Univerx builds AI on, and the controls that put them to work in AI agents, Univerx Assist and Semantic².
Trust & EuropeGDPR
Processing under Art. 28 GDPR, documents for your data protection impact assessment, retention and deletion policies and role rights. What exists you receive on request; what is a goal we call a goal.
Trust & EuropeSecurity & availability
Encryption in transit and at rest, identity and access management, network and API security, multi-zone availability and continuity plans: what Univerx is designed for so your customer service keeps running.
Guides2 November 2026 · 11 min readIntroducing AI agents responsibly
Disclosure, handover, permissions, logging, approval and measurement: what should be settled before an AI agent goes live in customer service – step by step, with a checklist for the project team.
Guides2 November 2026 · 11 min readRunning a contact centre in Europe
What the GDPR, the EU AI Act and German co-determination mean for routing, recording and analytics – and how to answer the questions from procurement, privacy and the works council before the first contact runs on a new platform.
Europe & regulation31 October 2026 · 5 min readThe EU AI Act in customer service: transparency duties, explained briefly
Article 50 of the AI Act is short, yet it affects almost every customer service operation that uses AI. What it says, and what follows for bots and assistance.
AI in customer service30 October 2026 · 5 min readDisclosing AI without breaking the conversation
A disclosure has to be clear, but it does not have to read like a package insert. A few rules for text, voice and the switch to a person.
AI in customer service30 October 2026 · 5 min readWhy the handover decides: AI agents and context
The most delicate moment in an automated conversation is not the hard question. It is the switch to a human. Plan it, and neither side has to repeat itself.
GlossaryEU AI Act
The EU's AI regulation classifies AI systems by risk and attaches duties to each class: prohibitions, strict requirements for high-risk AI, disclosure in dialogue. What that means for AI in customer service.
GlossaryAI agent
An AI agent is software that understands a request in natural language, acts in connected systems and hands over to a person when a decision or empathy is needed.
GlossaryAgentic AI
Agentic AI receives a goal instead of a single instruction. It breaks the goal into steps, uses approved tools and checks interim results. Its value grows, and so does the need for limits and oversight.
GlossaryHallucination
A hallucination is an AI answer that sounds convincing but is wrong or invented. It follows from how language models work and can be limited, but never ruled out completely.
See Univerx on your own example.
In a demo we show the platform along your channels and requests, from AI agents to video, workforce scheduling and analytics.
Online, in German or English, with no preparation on your side.