Legal
Safety & Crisis Protocol
Last updated: November 13, 2026 · Escape Enterprises LLC
Our promise. When someone may be at risk, Lucille stops and shows real crisis help. That message is written by our team in advance; it is never generated by AI.
1. If you need help now
In an emergency, call your local emergency number (911 in the US and Canada, 000 in Australia, 111 in New Zealand, 999 in the UK, Malaysia and Hong Kong, 995 in Singapore, 112 in India and Ireland).
- US and Canada: call or text 988
- UK and Ireland: Samaritans, 116 123
- Australia: Lifeline, 13 11 14 · New Zealand: call or text 1737
- Singapore: Samaritans of Singapore, 1767 · Philippines: NCMH, 1553
- Malaysia: Befrienders KL, 03-7627 2929 · Hong Kong: 18111 · India: Tele-MANAS, 14416
- Anywhere else: findahelpline.com
Lucille is an AI. She can't call for help, and she isn't a crisis service.
2. How we notice someone may be at risk
- On your phone: before a typed message leaves your device, Escape checks it for signs of suicidal thoughts, self-harm or wanting to die, including common ways people say it. When you talk to Lucille, the phone runs the same check on the transcript of what you said.
- On our servers: every message to Lucille is checked again before Lucille's AI writes a reply, for signs of suicide, self-harm and threats of violence. A spoken message is first turned into text by OpenAI, which transcribes voice for us, and that text is checked before any reply. Messages in other languages also go through a multilingual safety model.
- Disordered eating: when a message shows signs of it, Lucille responds gently, avoids numbers like calories or weight, and points to support.
- The checks are tested against a list of real-world example phrases, in English and other languages, and updated when we find gaps.
3. What happens next
- Lucille does not reply with AI. Instead, Escape shows a fixed message written by our team, with the crisis lines for your country, your local emergency number and findahelpline.com, and one-tap buttons to call or text.
- The "I'm really struggling" button on the check-in screen opens the same help at any time.
- We record that crisis resources were shown (not the words of the conversation in any report) so we can count referrals and improve.
4. Reminders that Lucille is an AI
- At the start of every conversation, typed or spoken.
- Every 3 hours (or about every 12 messages) during a conversation.
- Lucille never claims to be human, to have feelings, or to be a therapist.
5. Adults only
Escape is for adults 18 and older. Sign-up stops for anyone under 18, and we delete any account we learn belongs to someone under 18.
6. Reporting a reply
Tap Report under any reply from Lucille, or email support@escapeapp.ai with the date and time. Our team reviews safety reports within 2 business days, and we use them to fix Lucille and our safety checks.
7. Yearly reporting
Starting July 1, 2027, we report each year to the California Office of Suicide Prevention, and where required to other states, how many times we showed crisis resources and the protocols on this page. Reports contain no personal information.
8. Contact
Safety questions: support@escapeapp.ai, Escape Enterprises LLC.
Questions about this policy?
Write to us for anything: data access, corrections, deletion or general questions.