Room 55

AI Ethics

Honest about what the AI can and can't do.

Room 55 uses AI to listen, reflect, and help you feel less alone. We want to be completely transparent about how it works, where its limits are, and how we make sure it doesn't cause harm.

What the AI companion can do

  • Listen without judgment and reflect back what you share
  • Help you feel heard and less alone
  • Ask gentle questions to help you clarify your own thinking
  • Surface vetted resources when you're ready for them
  • Recognize signs of crisis and surface emergency resources
  • Maintain a consistent, calm, non-judgmental tone

What the AI companion cannot do

  • Diagnose mental health conditions
  • Provide therapy or clinical treatment
  • Prescribe or recommend medication
  • Contact emergency services on your behalf
  • Guarantee accurate information in all cases
  • Replace human connection or professional care

Human oversight

Right now, Room 55 does not have a human moderation team. The AI companion is the only layer of safety response in the platform. When the companion detects a potential crisis, it surfaces crisis resources (988, 911, Crisis Text Line) and encourages the person to reach out to a professional. It does not escalate to a human reviewer, because no such reviewer exists yet. We are telling you this plainly because you deserve to know what you are relying on.

If you believe the AI companion said something harmful or inappropriate, please email [email protected]. We read every message.

How the AI is trained

Room 55's AI companion is built on a large language model and guided by internally developed content principles. It is designed to follow trauma-informed communication practices: non-judgmental language, active listening, and avoiding unsolicited advice. It is not trained on your personal conversations.

Known limitations

  • The AI may occasionally misread tone or context.
  • It may not always recognize cultural nuances.
  • It is not infallible in crisis detection.
  • It cannot verify the accuracy of everything it says.

We are honest about these limitations because we believe transparency is part of safety. If something feels wrong, trust your instincts and seek human support.