AI Disclosure
Lily is an AI
Every OpenAlice widget identifies itself as an AI assistant, visibly and from the first message. We build to the transparency duties of Art. 50 EU AI Act: people must know when they are talking to a machine — the AI label is part of the product and may not be removed by customers.
What Lily is — and is not
- She answers from the customer's knowledge base, is instructed to say "I don't know" and to escalate to a human rather than invent.
- She does not make decisions with legal or similarly significant effect on you (no credit scoring, no eligibility decisions).
- No emotion recognition, no biometric categorisation, no social scoring — we do not build prohibited-practice features (Art. 5 EU AI Act).
- Voice output is synthetic. Where a cloned voice is used, the customer must hold the speaker's consent.
Human in the loop
Customers can enable live human takeover: a person joins the same conversation, visibly labelled. Every takeover is logged in the audit trail. Where the customer has enabled escalation, you can always ask for a human.
Our own operation
OpenAlice itself is operated by one human and many AI agents — the same transparency applies: our agents work under human oversight with safety gates, audit trails and a standing operating rule: Safety First. Science Second.
Provider transparency
Language models and voice run on the providers listed in the Privacy Policy § 5 — EU by default, US only with explicit visitor consent. Model choice per widget is documented in the customer dashboard.
Questions
Write to nao@openalice.eu — we answer transparency questions from anyone, customer or visitor.