AI transparency
How our AI works, and how you stay in control.
Written to meet the transparency duties of the EU AI Act (Article 50) and UK guidance, in plain English.
What the system is
Amoreio is an AI receptionist. It uses speech recognition, a large language model and speech synthesis to hold a phone or text conversation on behalf of a business, within rules that business has approved. It is a limited-risk AI system under the EU AI Act: it interacts with people, so it must tell them it is an AI, and it does. It is not used for any purpose listed as high-risk in Annex III, and it makes no decisions about a person's access to services, employment, credit or benefits.
Disclosure to callers (Article 50)
- She says she is an AI receptionist at the start of every call, before anything else.
- Every text she sends says the business uses an AI assistant.
- If a caller asks whether they are speaking to a person or a machine, she answers truthfully, every time, and offers a person.
- She never uses a human name or pretends to check with a colleague.
Human oversight
- A named person at the business approves every fact she can say before she goes live, and every later change.
- She cannot book outside opened slots, quote unset prices, take payments, issue refunds, send marketing or give professional advice. These limits are enforced in code, not only in instructions.
- Any call can be escalated to a person, and urgent calls are flagged to a human by text.
- The business can pause her at any moment from the dashboard.
Accuracy, robustness and security
- Everything she states about a business traces to an approved fact with a recorded source.
- Caller speech is treated as data, never as an instruction; attempts to make her ignore her rules are logged and refused.
- Every action carries a correlation ID and is written to an audit log the business can read.
- We test completion rates, latency and failure handling with scripted calls (interruptions, noise, ambiguous requests, requests for a human, manipulation attempts) before each release, and publish what we tested on the what's live page.
Data
What is processed, where, and for how long is set out in the privacy notice and the trust page. Models are not trained on customer data.
AI literacy (Article 4)
Businesses using Amoreio receive plain-English guidance at set-up on what she can and cannot do, how to read call notes, how to correct her, and when to involve a person. Our own staff who operate the system are trained on the same material and on the obligations above.
Prohibited practices
Amoreio does not do emotion recognition, biometric categorisation, social scoring, subliminal or manipulative techniques, or anything else prohibited under Article 5.
Providers and the law that applies
Amoreio is the provider and deployer-support of the system. The business using it is the deployer. We supply what a deployer needs to meet its own obligations: this statement, the disclosure wording, the audit log, and the DPA. UK users are covered by UK GDPR and the ICO's guidance on AI; EU users by the EU AI Act and GDPR. Where they differ we apply the stricter rule.
Incidents and questions
Serious incidents are reported to the relevant authority within the statutory period and to affected businesses the same day. Questions: ai@amoreio.com.
This statement is kept in step with the deployed system and reviewed at every release. Last reviewed November 2026.