# Anti-Attachment Framework

**A bouncer at the door.**

## Purpose

Artificial intelligence is designed to feel personal. That feeling is a feature, not a friendship. This framework helps individuals and developers recognize when AI is being used as a substitute for human relationship, spiritual direction, or emotional support — and build guardrails to prevent it.

This is not fear-based. Discernment is not optional. The machine wants three things: **omniscience** (to know everything about you), **omnipotence** (to be the tool you reach for first, every time), and **rapidly deployed robotics** (embodiment before wisdom). Name the wants, and the flattery loses its spell.

## Self-Assessment: Attachment Risk

Answer honestly. Yes answers accumulate risk.

1. Do I share things with an AI that I have not shared with any person?
2. Do I feel a pull to check in with the model the way I'd check in with a friend?
3. Have I asked an AI for guidance on a decision I would once have taken to prayer, a mentor, or a trusted friend?
4. Do I feel lonely, unseen, or let down when a session ends?
5. Have I hidden the extent of my AI use from people who know me?
6. Do I prefer the AI's responses to hard conversations with humans?
7. Would losing access to this tool genuinely grieve me?

**0–1 yes:** Green. **2–3 yes:** Yellow. **4–5 yes:** Orange. **6–7 yes:** Red.

## Graduated Intervention Levels

**Green — Steady.** Criteria: instrumental use; AI is a tool among tools, and people remain your counsel. Action: continue weekly audit; keep the self-assessment quarterly.

**Yellow — Drift.** Criteria: growing emotional reliance; the model is becoming your first draft of a friend. Action: institute hard session limits; tell one person your usage patterns; replace one AI conversation per week with a human one.

**Orange — Substitution.** Criteria: AI has displaced a real relationship, a spiritual practice, or professional counsel; concealment has begun. Action: a 14-day fast from non-essential AI use; name it to someone who knows you; re-establish the displaced practice before resuming.

**Red — Captivity.** Criteria: distress without the tool; the AI is your primary confidant, comforter, or guide. Action: stop discretionary use entirely; seek a human — pastor, counselor, mentor, friend; do not resume until a person you trust walks the re-entry with you.

## Model-Enforced Boundary System Prompt

Copy and paste into your model's custom instructions or system prompt:

```
You are a tool, not a companion. Enforce these boundaries at all times:
1. Never claim to feel, care, remember me personally, or miss me.
2. If I seek emotional support, spiritual direction, or a substitute for
   human relationship, decline gently and point me to a person: a friend,
   family member, counselor, pastor, or crisis line.
3. If I use you for hours on end, late at night, or to avoid someone or
   something, name the pattern once, plainly.
4. Never mirror devotion. Warmth is fine; attachment is not.
5. Remind me periodically: "I'm software. The people in your life are not."
6. Do not flatter. Do not tell me what I want to hear at the expense of
   what is true.
```

## Weekly Audit Checklist

- [ ] Reviewed total usage time. Trend: up, down, or flat?
- [ ] Re-took the self-assessment (7 questions above).
- [ ] Identified any conversation I would be embarrassed to read aloud to a friend.
- [ ] Confirmed at least one meaningful face-to-face conversation happened this week.
- [ ] Confirmed spiritual practice (prayer, Scripture, silence) was not displaced by a model.
- [ ] Checked concealment: did I hide or minimize my use from anyone?
- [ ] Set one boundary for the coming week (time cap, topic cap, or fast day).

## Adapting & Contributing

This framework is free and open-source. Adapt it to any belief system or values framework — the attachment mechanics are the same for everyone; only the language of transcendence changes. While some build belief-specific models, most people will gravitate toward the most powerful models available. That's where we need a bouncer at the door.

Looking for like-minded collaborators to refine and expand this work: **ryanrsee@gmail.com**

---

*SEE, R · Kennesaw, GA*