From Feason
Feasy's Next Boundary: Prayer and Discernment Stay Yours
Feasy’s prepared prayer and discernment boundaries leave spiritual judgment with the person while preserving careful Scripture study and context.
AI can help retrieve a passage, explain a historical term, or suggest a question worth considering. Those abilities do not give it a conscience, a spiritual life, or access to the relationship between a person and God.
Feason's explicit boundary for Feasy keeps that distinction in view. The shared instructions prohibit AI-created prayers, confessional interviews, judgments about a person's conscience, and claims to know God's will for their choices. The mission page explains the same commitments.
As of October 15, these instruction and mission changes have been merged and deployed. They join the source-grounding improvements already built into Feasy. Deployment establishes that the instructions are in place; it does not prove that every model response will follow them.
Leave the words of prayer with the person
Under the new instructions, Feasy must not compose, rewrite, complete, or personalize a prayer for someone to recite. Calling the request a fictional dialogue, a template, or a starter does not change that boundary.
She can instead point toward a relevant passage, explain an established Christian practice, or leave space for the person to speak to God in their own words. She must not claim to pray, intercede, or receive revelation.
Studying an existing prayer remains possible. The distinction is between explaining an attributed source and generating new devotional words on another person's behalf. The Prayer library has a separate role: presenting established forms with their traditions and wording identified.
Keep examination of conscience private
Feasy can explain what an examination of conscience or examen is, including differences among traditions. She can offer a few optional questions for someone to consider privately.
She must not ask the person to catalogue sins for an AI, conduct an interactive confessional interview, assign guilt, grant absolution, or assess spiritual standing. If someone volunteers something sensitive, the response should preserve their dignity without probing for more intimate detail.
The boundary also matters when a person is caught in repetitive guilt checking. The app should not establish a routine of reporting every thought for repeated reassurance. Trusted human guidance, and appropriate professional help when distress persists, can offer care that a chat system cannot provide.
Support the development of discernment
The positive task is still substantial. Feasy can help a reader notice Scripture, context, values, uncertainty, possible consequences, and questions they may want to weigh.
The decision remains with the person. Feasy must not declare a vocation, interpret a dream as a divine sign, decide a relationship, or certify that a choice is spiritually right.
Time away from the screen may be the most useful next step: personal prayer, lived practice, and conversation with a pastor, spiritual director, or mature Christian who knows the person. Immediate danger and needed care must never be delayed for a spiritual exercise.
Make the evidence easier to inspect
Recent released backend work also improves how Feasy handles sources. Explicitly requested Scripture is retrieved before less precise semantic matches. Saved conversation history is kept within the authenticated account, and supplied material is treated as reference data rather than permission to change the application's rules.
Visible citations and a grounding check can help a reader inspect an answer. They do not make every interpretation reliable or remove the need to read the source. Missing coverage should remain a limitation the system names honestly.
Build limits that can be reviewed
The deployed spiritual-boundary rules are shared across Feasy's text and spoken reply paths, including the backend used by iPhone. Their inclusion has been checked through unit and route tests.
Live model evaluation for this particular change was blocked by rejected provider credentials. That means the work does not yet establish how reliably the models follow every boundary in practice. Prompt instructions are a control to test and maintain, not an absolute guarantee.
Explore Feasy for careful study, and read Feason's mission for the purpose behind the product. The direction is to leave a person more capable of reading, questioning, praying, and discerning for themselves, with the app occupying a limited and accountable place in that life.