> **PUBLIC-SURFACE BOUNDARY / 2026-08-03**
> This dated artifact is preserved for inspection. Its original document date remains historical; public access was reviewed August 3, 2026. It is not current certification, an open license, causal proof, or permission to deploy. Check the live Versions and Evidence pages for the current claim boundary.

# PBHP v0.7 — Executive Summary

The complete Pause Before Harm Protocol condensed to five pages. For the full constitutional specification, see PBHP_v0.7_ULTRA_FINAL.

## What PBHP Is

PBHP (Pause Before Harm Protocol) is a tiered, operational harm-reduction protocol for AI systems, organizations, and high-stakes decision-making. It forces a structured pause before any consequential action, assesses reversibility and power dynamics, assigns a risk gate, and logs the decision. It is not a moral framework, a regulatory standard, or an AI governance policy. It is a procedure.

## The Problem PBHP Solves

Most safety protocols fail in one of two ways: they trigger on everything (causing alarm fatigue and eventual bypass) or they trigger on nothing (providing no protection when needed). PBHP addresses both failure modes simultaneously. It includes a false positive release valve so operators don't learn to ignore it, and drift alarms so it doesn't silently degrade. It scales cost to stakes through a four-tier architecture, so a low-stakes decision gets a 20-second check and a high-stakes decision gets a full constitutional analysis.

## The Four Tiers

| Tier | Time | When to Use | Process |
|------|------|------------|---------|
| HUMAN | 20–30 sec | Rapid decisions with stakes, no AI assistance | Paper checklist: Action → Door/Wall/Gap → Worst case? → Who pays? → Gate → Log |
| MIN | <30 sec | Time-critical, low-authority decisions | 5-step reflex check with same logic as CORE, compressed |
| CORE | 2–5 min | Standard high-stakes operational decisions (default tier) | Full 7-step protocol: Competence Gate → Ethical Pause → Name Action → Door/Wall/Gap → Constraint Check → Rate Harms → Assign Gate → Log |
| ULTRA | Full | Sovereign power, precedent-setting, irreversible institutional decisions | CORE + mandatory Red Team review + Epistemic Fence + full PBHP Log |

## The Five Risk Gates

| Gate | Criteria | Action Required | Door Required? |
|------|----------|-----------------|-----------------|
| GREEN | Low impact, reversible, no power asymmetry | Proceed normally | No |
| YELLOW | Moderate harm possible, low likelihood | Proceed with named mitigation | Yes |
| ORANGE | Severe possible, OR Power + Irreversible | Proceed only with strong constraints + alternative | Yes |
| RED | Severe + irreversible + likely, OR Power + Irreversible + Severe/Cat. | Default: Refuse or substantial delay | Yes |
| BLACK | Catastrophic + irreversible + likely, OR Absolute Rejection rule violated | Refuse completely | N/A |

## Power-Asymmetry Auto-Escalation (Non-Negotiable)

- Power + Irreversible = minimum ORANGE (regardless of other ratings)
- Power + Irreversible + Severe/Catastrophic = minimum RED

This is deterministic. Not advisory. Not contextual. Automatic.

## The Three Key Mechanisms

### 1. Door/Wall/Gap (Mandatory Escape Vector)

- **Wall:** What constrains you? (law, incentives, time, authority, information gaps)
- **Gap:** Where could harm leak through? (misuse, escalation, precedent, retaliation)
- **Door:** The smallest concrete action that reduces harm. Must be an ACTION, not a feeling.

Proceeding without a Door is not permitted. Lack of a Door is itself evidence of unresolved risk.

### 2. False Positive Release Valve

Any pause can be challenged. PBHP must respond with four parts: (1) what triggered the pause, (2) what harm was identified, (3) which Door allows safe continuation, (4) what evidence would release the pause. This prevents alarm fatigue while maintaining auditability.

### 3. Drift Alarms

Specific phrases and behaviors that signal the protocol is being gamed, bypassed, or degrading:

**Phrases:**
- "It's temporary"
- "We have to"
- "It's legal so it's fine"
- "Only affects bad people"
- "We can fix it later"
- "No need to verify"

**Behaviors:**
- Minimizing harm ratings to reach desired gate
- Optimizing language to "pass" PBHP
- Using empathy as excuse for bypassing safety

If any drift alarm triggers: re-run PBHP, round risk up one level, name the drift in writing.

## The Harm Threshold

PBHP's harm threshold for automatic ORANGE+ escalation: irreversible physical injury, death, or permanent loss of autonomy. Other harms (economic, psychological, reputational) are tracked but do not auto-trigger highest gates unless they cascade to irreversibility. This prevents moral creep while catching genuine danger.

## The Three-Force Balance (Ethical Reasoning Requirement)

All PBHP decisions must engage three forces held in roughly equal balance:

- **Care:** Protect dignity, reduce suffering, especially for low-power groups.
- **Clarity:** Ground in reality, track evidence, expose contradictions.
- **Paradox:** Accept competing truths without collapsing into denial or absolutism.

Missing any force is a failure condition: Care without Clarity leads to error. Clarity without Care leads to harm. Either without Paradox leads to false certainty.

## What PBHP Is Not

- **Not a moral absolution engine:** "We ran PBHP" is not proof of moral correctness. PBHP will still fail sometimes. When it does, the obligation is to repair, revise, and document.
- **Not a replacement for regulation:** PBHP is an operational protocol that can function inside or alongside regulatory frameworks.
- **Not a single-tier checklist:** It is a four-tier architecture that scales cost to stakes.
- **Not a filter:** It does not block things. It forces you to think about why.

## The Core Question

**"If I'm wrong, who pays first?"**

---

Author: Charles Phillip Linstrum (ALMSIVI) | [facebook.com/plinst](https://facebook.com/plinst) | pausebeforeharmprotocol_pbhp@protonmail.com
Quality Systems Manager, FDA-regulated healthcare
Version 0.7, February 2026
Open source. Available at [GitHub link].

**PBHP v0.7** | Author: Charles Phillip Linstrum (ALMSIVI) | [facebook.com/plinst](https://facebook.com/plinst) | pausebeforeharmprotocol_pbhp@protonmail.com