1. Statutory Foundation & Governance
Tovelu provides this AI Transparency, Explainability & Ethics Statement pursuant to:
- Article 50 of the European Union Artificial Intelligence Act (Regulation (EU) 2024/1689): Transparency obligations for AI systems interacting with natural persons or generating content;
- United States Federal Trade Commission (FTC) AI Directives: Ensuring representations regarding artificial intelligence capabilities, accuracy, and training data are truthful and scientifically substantiated;
- Tovelu Master Constitution (Section 5): The supreme company doctrine of Deterministic Over Generative and The Explanation Standard.
2. Dual-Engine Architecture: Deterministic vs. Generative Systems
To prevent AI hallucination in health pathways, Tovelu maintains a strict architectural separation between Deterministic Algorithmic Systems and Generative AI Systems:
- Vitality Score calculation (0–100)
- Pillar sub-scores (Physical, Metabolic, Sleep, Mental)
- Daily hydration volume ranges based on weight/climate
- Habit milestone & 28-day streak verification
- Written in deterministic TypeScript/Python code
- Zero probabilistic token prediction
- Natural language health report summaries
- Personalized educational habit explanations
- User habit adherence encouragement
- Formatting habit reminders and notifications
- Constrained by Retrieval-Augmented System (RAG)
- Strict prompt safety rails and verification gates
3. The Tovelu Explanation Standard
Under Section 6 of the Tovelu Master Constitution, every health recommendation generated or assisted by artificial intelligence must satisfy the five-part Explanation Standard:
4. Prohibited AI Behaviors & Emergency Override
Our system prompts and automated filters enforce hard boundaries. Tovelu's AI is strictly forbidden to:
- Generate medical diagnoses or name specific diseases as a clinical finding;
- Recommend starting, stopping, or altering prescription medications;
- Provide clinical drug dosage calculations or therapeutic titration schedules;
- Attempt to provide psychotherapy, psychological crisis counseling, or suicide intervention.
If an interaction detects keywords associated with self-harm, suicide, or acute medical emergencies, the AI engine is immediately bypassed, and the platform displays our hardcoded Emergency Crisis Protocol with direct emergency phone numbers (988, 111, 112, 14416).