EU AI Act statement

Regulation (EU) 2024/1689 (the “AI Act”) entered into force on 1 August 2024. Its prohibitions have applied since 2 February 2025, the obligations for general-purpose AI models since 2 August 2025, and the transparency duties of Article 50 — the ones that matter for a customer-service chatbot — since 2 August 2026. This statement describes how CYBIX and its AI agent BIX meet them, and which duties sit with you as the deployer.

BIX is an AI system that interacts with natural persons: it answers website visitors and customers, drafts replies for agents, summarises and translates conversations. That places it under the transparency obligations of Article 50. It is not a prohibited practice under Article 5 — it does no social scoring, no emotion recognition, no biometric categorisation and no manipulative or subliminal techniques — and, used for customer support, it is not a high-risk system under Annex III.

If you deploy BIX in an Annex III context (for example screening job applicants, deciding on credit, insurance or access to essential services, or in education or law enforcement), the high-risk regime applies to that use and you must not do so without first agreeing an appropriate configuration with us.

• Visitors are told they are talking to an AI: the messenger labels BIX as an “AI Agent”, the default greeting says so, and the label is part of the product, not a setting.
• When a human takes over, the header switches to your brand or the agent, so the person always knows whether a human or the AI is answering.
• Suggested replies and drafts shown to your team are marked as AI-generated inside the dashboard before an agent sends them under their own name.
• BIX does not generate synthetic audio, images or video.

• A visitor can ask for a human at any point; your team can take over any conversation, which pauses BIX for that thread.
• You decide whether BIX answers autonomously, only drafts, or is off per workspace and per widget.
• Every BIX answer is stored in the conversation history, so you can review it, correct the knowledge base and see exactly what was said.
• Answers are grounded in your own approved content; BIX is instructed not to invent policies, prices or commitments.

BIX runs on third-party general-purpose AI models reached through the providers on the sub-processor list. The obligations for those models (technical documentation, copyright policy, training-data summaries) rest with their providers. CYBIX sends only the conversation text and retrieved knowledge needed for a reply, and does not use your data to train models.

• Keep the AI disclosure visible: do not rename BIX in a way that presents it as a person, and do not remove the “AI Agent” label through custom code.
• Give the people who configure and supervise BIX enough AI literacy (Article 4) to understand what it can and cannot do; our help center courses cover this.
• Inform your customers in your own privacy notice that support may be handled by an AI system, and how to reach a human.
• Do not feed BIX instructions that would make it discriminate, manipulate or profile people in ways the AI Act or GDPR forbid.
• Tell us if you intend a high-risk use so the configuration, logging and documentation can be adjusted first.

Serious incidents involving BIX — harmful advice acted upon, discriminatory output, or a disclosure failure — are handled through the same incident process as data breaches: report to support@cybix.app, we investigate, fix and inform the affected customers. Questions about this statement go to the same address.