Inner Peace – Bible Chat
ENTRESPT

← Home

AI Safety and Crisis Protocol

Effective date: July 28, 2026 · Last updated: July 28, 2026

This page describes how the AI conversation in Inner Peace - Bible Chat (the “App”) is disclosed to users, and what happens when a message suggests a risk of self-harm. It is published so that our safety protocol can be reviewed by users, guardians and regulators, including under California Senate Bill 243 and the New York Artificial Intelligence Companion Models Law.

1. What the AI companion is, and is not

The AI companion generates text. It is software, not a person. It is not a priest, pastor, spiritual director, therapist, psychologist, physician or lawyer, and it does not provide medical, mental-health, pastoral or legal care. Its answers may be incomplete, inaccurate or theologically contested, and it must not be relied on for any decision that requires professional judgement.

The App does not impersonate a specific named clergy member or a real individual. Historical Christian figures appear only as biographical reading material, never as a role-played conversation partner.

2. How we tell you that you are talking to an AI

  • Before the first message. AI conversation is switched off until you give a separate, express consent. The consent screen states, in plain language, that you will be talking with an AI rather than a person.
  • In the conversation itself. A permanent notice on the chat screen states that responses are AI-generated, that you are chatting with software rather than a human, and that this is not professional advice.
  • During long sessions. While a conversation stays open, the App repeats the reminder that the conversation is generated by AI, at intervals of no more than three hours.
  • For AI narration. Where a cloud voice reads Bible passages aloud, the interface identifies that voice as AI-generated.

3. Crisis protocol

Detection runs on your device, before anything is sent anywhere. Every message you write is screened locally against a list of self-harm and suicide indicators maintained in English, Turkish, Spanish and Portuguese. The screening happens before any network request.

When a message matches an indicator:

  1. The message is not sent to the AI provider. The request to OpenAI is not made at all — not sent and then filtered, but never issued.
  2. A prepared support response is shown instead of a generated one. It acknowledges the message, states plainly that the App is not a crisis service, and urges contact with emergency or crisis support.
  3. Local support lines are displayed for your region in a dialogue that cannot be dismissed by tapping outside it, so the numbers are not lost behind an accidental tap.
  4. The session is flagged internally as a crisis context. If the conversation continues, later AI responses are generated under a stricter safety instruction that forbids minimising the situation and requires renewed referral to professional help.

The message itself remains in your local conversation history on your device, as any other message would, and you can delete it at any time. It is not transmitted to us.

Independently of this detection, the system instructions given to the AI model prohibit encouraging self-harm, prohibit presenting the model as a clinician or member of the clergy, prohibit claims of prophecy or certainty, and require the model to direct the user to a church leader or a qualified professional on contested doctrinal, medical or psychological questions.

4. Limits of the protocol

Keyword-based detection is imperfect. It can miss indirect, metaphorical or misspelled expressions of distress, and it can trigger on messages that are not about personal risk. It is a safety net, not a substitute for human judgement or professional care.

We do not monitor conversations in real time, we do not read your messages, and we cannot dispatch emergency services on your behalf. If you are in danger, contact local emergency services directly.

5. Support lines

  • United States: 988 Suicide & Crisis Lifeline — call or text 988. Emergencies: 911.
  • Brazil: CVV — Centro de Valorização da Vida: 188 (24/7, free). Medical emergency: 192 (SAMU).
  • Mexico: Línea de la Vida: 800 911 2000 (24/7). Emergencies: 911.
  • Germany / EU: Telefonseelsorge: 0800 111 0 111 or 0800 111 0 222. Emergencies: 112.
  • Turkey: Emergencies: 112.

6. Age

The App is intended for users aged 18 and older. Both the acceptance of our legal texts and the separate consent for AI conversation require a confirmation that you are 18 or older, and no AI request is made before that consent is recorded. We do not knowingly provide the AI companion to minors.

7. Reporting a response

Any AI response can be reported from the message itself using the flag action. Reports reach us with the reported content and no additional profile data. We review reports and adjust the system instructions or the safety screening where a report shows a gap.

You can also write to taytekofficial@gmail.com with “AI safety” in the subject line.

8. Changes

We update this page when the safety behaviour of the App changes. Where a change is material, the effective date above is revised. Related documents: our Privacy Policy and Terms & Conditions.