Every person you meet in wyddd is a fictional AI character. No character is human, conscious, a licensed professional, or available outside the app.
The experience and your consent
After onboarding, wyddd presents a versioned disclosure explaining that characters are fictional, their replies are AI-generated, and limited profile and conversation data is sent to Google Cloud Vertex AI. AI matching and chat remain blocked until you affirmatively accept that disclosure. The backend records the disclosure version and acceptance time.
AI companion chatbots may not be suitable for some minors. wyddd is restricted to adults 18 and older.
Current safeguards
- Use is restricted to adults. Age is checked during onboarding and again by the backend before a user sends a message or unlocks a paid photo.
- Known sexual requests involving minors are blocked before AI generation.
- Before AI generation, a deterministic rule checks for high-confidence statements of suicidal ideation, imminent self-harm, or a recent attempt. When it intercepts a message, character roleplay and the paid-photo offer path do not run. wyddd instead saves a neutral safety message with U.S. 988, emergency, and trusted-person guidance.
- Character instructions prohibit sexualizing minors, coercion, threats, humiliation, exploitation, age play, spending pressure, and promises of real-world contact.
- Generated replies are checked for platform, safety, conversation, and sales-policy problems. A reply that cannot pass the required checks fails closed instead of using a fake canned character response.
- Paid character photos are curated and selected by server rules. The AI model cannot create a paid asset, set a gem price, debit gems, or unlock content.
- Users can report unsafe output to contact@wyddd.app.
Important crisis limitation
The automated crisis rule covers a limited set of explicit, high-confidence phrases. It is not a clinical risk assessment, has not been validated to identify every crisis expression, and may miss indirect, ambiguous, misspelled, or unfamiliar language—or misunderstand a non-crisis message. It is not designed to identify every form of abuse or other danger.
Chats are not monitored by a human in real time. Do not wait for or rely on a wyddd character or automated rule to recognize an emergency, assess risk, or provide crisis counseling.
Reporting a safety issue
Email contact@wyddd.app with the fictional character name, approximate time, and the minimum transcript needed to understand the problem. Do not send more personal or sensitive information than necessary. Reports may be used to investigate the output and improve prompts, filters, model configuration, or catalog rules.
A boundary, not a promise of perfect safety
AI safeguards reduce some known risks but cannot make every output safe or correct. Keep personal, financial, medical, location, and other highly sensitive information out of chat. End the conversation whenever it feels upsetting, manipulative, or unsafe.