AI Transparency

ArQonnect AI replies are grounded in FAQs, documents, and policies uploaded by the business customer. The system retrieves relevant passages and asks an LLM to draft a reply. It does not claim unrestricted general knowledge or capabilities that are not shipped.

1. First-interaction disclosure

Where EU AI Act transparency rules apply (and as a product default), the first AI reply in a conversation includes a clear notice that the assistant is automated, unless the AI nature is already obvious from the channel context. A statement on this page alone is not a substitute for in-product notice.

2. Human escalation

End-users can request a person. Low-confidence, complaint, sensitive, or unknown queries are handed to the business customer's human inbox when configured. Agent availability is controlled by the customer.

3. Limitations

• Answers may be incorrect or outdated if knowledge is stale.
• Pricing, medical, legal, and financial claims must stay within approved knowledge and lawful customer workflows.
• ArQonnect does not offer voice cloning of real people, deepfakes, or deceptive impersonation tools.

4. Data handling

Prompts may include the end-user message, retrieved knowledge chunks, and limited conversation context. LLM vendors and retention are described in the Privacy Policy. Meta Business Solution Data is not used to train or improve shared / general models where prohibited. Tenant data is isolated.

5. Contact

Questions: Compliance@arqonnect.io