safety and limitations.
AMYY is an AI companion, not a person, therapist, or emergency service. This page explains the safety features in the current beta and where they can fall short.
if you need urgent help.
For immediate danger, contact your local emergency service directly. AMYY cannot contact emergency responders or someone you trust on your behalf. Conversations are not monitored by a live crisis team.
For human support, find a helpline in your country or see the crisis resources. You can use these resources without waiting for AMYY to flag a message.
automated checks.
The app checks submitted messages for configured risk-language patterns. During model-generated replies, it also reads the model's risk classification when available. Either a matching escalation pattern or a high-risk model classification can cause the app to show crisis resources.
Some detected patterns receive a prepared support response before a model request. Other signals are evaluated as the model reply is processed. These checks can miss a crisis or flag something that is not a crisis, especially when language or context is unclear. They are engineering safeguards, not a clinical assessment or a clinically validated screening tool.
when a message is flagged.
AMYY aims to respond supportively and show human-support resources, including a worldwide helpline directory and regional options. The response may encourage reaching out to someone you trust or contacting emergency services directly. AI-generated replies and automated checks can still be wrong, delayed, or unavailable.
A flagged turn can create a separate safety record containing a message hash, timestamp, and category. The original message remains in the conversation history. See the privacy policy for storage, retention, and deletion details.
internal beta review.
Authorized AMYY team members can review stored conversations and safety records to investigate problems and improve the service. This access does not mean each flagged message is reviewed, that a clinician is involved, or that someone will respond within a particular time. Do not use the app or email as an emergency contact.
AI disclosures and break reminders.
Onboarding identifies AMYY as AI and explains that it is not a therapist. The chat interface also identifies AMYY as an AI companion.
During continuing conversation, the server tracks an approximately three-hour reminder interval for all users. A later completed chat reply can ask the interface to show a break reminder. This is triggered by chat activity, not an alarm that runs continuously while a page is open; outages or interrupted responses can prevent a reminder from appearing. You can leave or take a break whenever you want.
questions or concerns.
For a non-urgent question about these features, email connect@obvix.io. For immediate danger or urgent emotional support, use the human-support resources above.