AI Transparency Statement

1. Our Commitment

MorVoice is committed to transparent, safe, and ethical use of Artificial Intelligence. We align with the EU AI Act, GDPR, and international best practices to ensure our AI systems are lawful, fair, explainable, secure, and subject to meaningful human oversight. Our AI assists humans — not replaces them — particularly in healthcare, communication, and voice technologies.

3. Intended Use, Prohibited Use

• Allowed: accessibility, customer support, productivity, content generation with disclosure, healthcare pre-communication and scheduling (non-diagnostic).
• Prohibited: deception, impersonation without consent, unlawful surveillance, discrimination, medical diagnosis, high-risk use without contractual authorization and safeguards.

Synthetic speech must be disclosed to end users where contextually required; medical scenarios remain assistive only.

4. Data Sources and Model Training

• No training by default: customer inputs/outputs (text, audio, metadata) are not used for training unless the customer opts in via written agreement.
• Training data categories: publicly available text/speech corpora with license compliance, curated vendor datasets, and synthetic data; personally identifiable content is excluded or anonymized.

5. Human Oversight and Control

All automated decisions remain under human supervision. Operators can pause, correct, or override outputs at any time. In healthcare contexts, AI outputs are assistive and never diagnostic.

6. Fairness, Bias, and Robustness

We regularly evaluate models using multilingual suites, demographic balance checks, counterfactual tests, and randomized audits. When needed, we commission independent reviews. Detected bias triggers mitigation (prompting/guardrails, data balancing, or model updates).

7. Explainability and Auditability

For every AI system we maintain internal documentation: model version and purpose, decision logic overview, processed data categories, reliability indicators. We keep audit logs (excluding sensitive content) to support traceability and lawful requests.

8. Content Provenance and Disclosure

Where technically feasible, we apply provenance measures (e.g., metadata/watermarks) to synthetic speech and clearly disclose AI-generated content in user interfaces or documentation. Unauthorized recording, reverse engineering, or replication of MorVoice Voices is prohibited.