AI Coaching Disclosure

1. You Are Interacting with an AI System

Nora's AI Coach is powered by artificial intelligence, specifically Anthropic's Claude language models. When you have a coaching conversation, you are communicating with an AI system, not a human.

2. What the AI Does

The AI system performs the following functions:

• Coaching conversations — Real-time dialogue about emotional intelligence, leadership, and professional development using Claude Sonnet.
• Session summaries — After sessions, Claude Haiku extracts key themes, emotional arcs, competencies discussed, commitments made, and open threads.
• Coaching profile (dossier) — Claude Haiku maintains a long-term profile that evolves as you use the platform, capturing patterns, strengths, growth areas, and active themes.
• EQ score inference — After substantial coaching sessions (4+ messages), Claude Haiku evaluates whether your competency scores should be adjusted based on behavioral evidence in the conversation.
• Safety classification — Claude Haiku asynchronously reviews conversations for safety concerns (crisis indicators, harmful content).
• Emotional pattern analysis — Your mood check-ins and emotional data are analyzed for patterns over time.

3. What the AI Is Not

The AI Coach is not:

• A licensed therapist, psychologist, or counselor
• A medical professional or diagnostic tool
• A substitute for professional mental health services
• A legal or financial advisor
• A crisis intervention service

If you are in crisis or experiencing a mental health emergency, please contact your local emergency services, a crisis helpline, or a qualified mental health professional immediately.

4. AI Models Used

These models are provided by Anthropic. Your data is sent to Anthropic's API for processing. Anthropic does not use customer API data to train their models.

5. Safeguards & Human Oversight

• Score change limits: AI-inferred EQ score adjustments are capped at ±0.2 per competency per session. If total drift from your last human assessment exceeds 1.0, automatic inference is suspended.
• Minimum conversation threshold: Score inference only triggers after sessions with at least 4 user messages.
• Safety review: A safety classification system flags concerning conversations for human review. Events are categorized by severity (low through critical).
• Encryption: All coaching conversations are encrypted with AES-256-GCM using per-user encryption keys. Messages are decrypted only in-memory during active sessions.
• Audit trail: All AI actions (scoring, profiling, safety events) are logged with full metadata.

6. AI Limitations

• The AI may produce inaccurate, incomplete, or biased responses
• AI-generated scores and profiles are approximations, not definitive assessments
• The AI does not have real-time knowledge of events or your physical state
• The AI cannot detect all safety concerns — it supplements but does not replace human judgment
• Coaching quality depends on the information you share; the AI works best with honest, detailed input

7. Your Control

• Consent: AI processing requires your explicit consent, which you provide during onboarding
• Withdrawal: You can withdraw AI processing consent at any time via Profile Settings. This will disable AI coaching features but will not delete existing data.
• Data export: You can export all your data, including AI-generated content, via Profile Settings
• Data deletion: You can request deletion of your account and all associated data

8. EU AI Act Compliance

In accordance with transparency obligations, we disclose that this AI system assesses psychological characteristics and generates profiles that may influence coaching recommendations. Users are informed of AI involvement at every stage. Human oversight mechanisms are in place, and users retain full control over their participation.

9. Contact

Questions about our AI practices: [email protected]