EU AI Act — Transparency Disclosure

Regulation (EU) 2024/1689 (the AI Act) — Edition: 2026-05-26

1. Provider identification

Provider of the AI system: Project Line, support@projectlineil.com.

2. AI components present in the Service

Voice receptionist — natural-language audio agent that answers calls for the clinic.
Drug-safety helper — surfaces potential interactions from the patient's stored medication list.
Form auto-fill — pre-fills routine appointment templates from the patient's history.
Clinical-decision suggestions — non-binding suggestions for follow-up (PHQ-9 / GAD-7 thresholds, refill timing).

3. Risk classification under Annex III

The Service is positioned as a documentation and workflow tool. Clinical-decision-support functions surface information to the licensed clinician for their independent review and decision. The treating clinician retains exclusive medical authority; the Service does not autonomously triage, diagnose, prescribe, or refuse care. The Service is not a medical device under Regulation (EU) 2017/745 (see the medical device status page), so it is not caught by Art. 6(1) read with Annex I of the AI Act. No Annex III use case applies either — in particular point 5(a), which covers AI systems used by or on behalf of public authorities to evaluate eligibility for essential public assistance benefits and services, and point 5(d), which covers the evaluation and classification of emergency calls and the dispatch of emergency services, including emergency healthcare patient triage: the Service performs no triage and decides no one's entitlement to care. On that basis the Service is not operated as a high-risk AI system.

If a clinic elects to use AI output to make autonomous clinical decisions, the clinic itself becomes the deployer of a high-risk AI system under Art. 26 AI Act and assumes the corresponding obligations. The Service Agreement prohibits this use.

4. Art. 50 transparency obligations met

Art. 50 applies from 2 August 2026 (Art. 113 AI Act); the notices below are already in place.
Phone. Patients calling the AI voice agent hear a spoken notice at the start of every call telling them they are speaking to an AI assistant (Art. 50(1)).
Calls to the AI voice agent are NOT recorded — the conversation is processed in real time (speech-to-speech) with no audio or transcript stored.
Synthetic voice output is identified as such on first contact.
SMS. Automatic replies carry a notice that the reply is automated and that a person at the clinic will see the patient's message.
WhatsApp. Only pre-approved template notifications are sent; no AI conversation is conducted on that channel.
In-app assistant. Intended for clinic staff and labeled as an assistant in the interface itself.
The clinic's medical records explicitly tag any AI-generated text in the appointment notes.

5. General-purpose AI model sources

Component	Provider	Model	Status
Voice + chat	OpenAI	GPT-Realtime family	GPAI provider — see openai.com/policies
Image OCR	OpenAI	GPT-Realtime family (vision)	GPAI provider

6. Quality management

Internal quality-management system covers data governance, log retention, change management, and post-market monitoring. Incidents are recorded in Project Line's internal log and material AI-related incidents are reported to clinic controllers within 72 hours.

7. Right to human intervention

Every AI output is reviewable and editable by the treating clinician. Patients always reach a human contact through the clinic; the AI voice agent transfers to a human on request.

8. Updates

This disclosure is updated whenever a new AI capability is added or a model provider changes. Material changes are announced under the sub-processor change procedure.