AI Transparency Notice

1. What AI does

At your request, MyBubble may send the minimum relevant content to an external AI provider to extract text, suggest meaning, answer from your own Bubbles or rank grounded sources. A private non-AI save path is available. Passive inferred profiling and long-term learning are off at launch.

2. Important limits

AI can make mistakes, omit context or misunderstand an original. Verify important dates, amounts, people, obligations and conclusions against the original. AI output is not professional medical, legal, tax or financial advice. You confirm consequential actions where offered. MyBubble is not an emergency service and must not be the only source for urgent decisions.

3. Provenance and marking

Generated output is visibly identified as AI-derived and carries durable structured origin and a versioned machine-readable marker in storage, API responses and exports where meaningful. Sources and user corrections keep distinct provenance. MyBubble does not present this proprietary marker as proof of a universal watermarking standard; provider and emerging interoperable technical-standard evidence remains part of ongoing compliance verification.

4. Sensitive data

Ordinary AI mode is not consent to process special-category data. One unselected account-level consent covers AI processing of sensitive information you intentionally provide about yourself for MyBubble's organisation and understanding features. It is not repeated for each Bubble and can be switched off in Settings.

AI processing of another person's sensitive data and criminal-conviction/offence data is off at launch. MyBubble does not add a new external classifier merely to determine the subject; if scope cannot be determined reliably, the AI processing is not treated as legally approved.

5. Withdrawal, correction and reporting

Withdraw a sensitive scope in Settings. Future provider calls stop immediately; linked AI-derived state is removed while private originals and your corrections remain. You can correct generated fields. You can report unsafe or incorrect Ask Bubble and Smart Search answers in the app; for other AI output, contact [email protected]. A withdrawn authorization never reactivates automatically.