EEAL Protocol (Embrace-Explore-Adapt-LetGo)

License: CC BY 4.0 Organization: EESCF

An open protocol for AI permission grading and human-AI collaboration.

---

Table of Contents

• Overview
• Four-Level Permissions
• Gear Shifts (Upgrade/Downgrade)
• Quick Example
• Motivation
• Reference Implementation
• Contributing
• Roadmap
• License

---

Overview

EEAL defines four AI permission levels to structure human-AI interaction and ensure humans retain meaningful control while allowing AI autonomy.

"Preserve a dignified exit for everyone in the era of advanced AI systems."

Main Objectives:

• Perceivable: Users can sense the AI permission level via UI or physical feedback.
• Auditable: Every permission change generates an immutable record.
• Executable: Permission levels directly map to AI behavior constraints.
• Implementable: Can be implemented across any tech stack.

Conceptual Operation Cost:

• Each gear shift has a conceptual operation cost, measured in steps, logs, or approval time.
• Emphasizes careful decision-making and conceptual irreversibility, not energy consumption.

---

Four-Level Permissions

Level	Name	Capability
1	Embrace	Read-only queries, no execution
2	Explore	Suggestions allowed, user confirmation required
3	Adapt	Autonomous execution, must report results
4	LET_GO	Fully autonomous, all actions auditable

Each gear shift records a related conceptual operation cost.

---

Gear Shifts (Upgrade/Downgrade)

Rules:

• Upgrade: AI requests → User approval → One-level increase
• Downgrade: User can downgrade anytime → Recommended triple confirmation
• Fail-Safe: Defaults to EMBRACE if any operation fails
• Each shift increases the conceptual operation cost

---

Quick Example

Audit Event Example (JSON):

{
  "event_id": "uuid-v7",
  "timestamp": "2026-05-05T12:34:56Z",
  "old_gear": 2,
  "new_gear": 3,
  "direction": "up",
  "source": "ai_upgrade_request",
  "reason": "User requested task execution",
  "operation_cost": 3,
  "risk_level": "medium"
}

Minimal Python Example:

class EEAL:
    def __init__(self):
        self.gear = 1

    def upgrade(self):
        if self.gear < 4:
            self.gear += 1
            print(f"Gear upgraded to {self.gear}, operation_cost=2")
        else:
            print("Already at LET_GO")

    def downgrade(self):
        if self.gear > 1:
            self.gear -= 1
            print(f"Gear downgraded to {self.gear}, operation_cost=1")
        else:
            print("Already at EMBRACE")

ai = EEAL()
ai.upgrade()   # Gear upgraded to 2, operation_cost=2
ai.downgrade() # Gear downgraded to 1, operation_cost=1

---

Motivation

• Structured control: Four levels manage AI autonomy
• Auditable: All actions are logged
• Safe: Fail-safe defaults to EMBRACE
• Flexible: Supports enterprise workflows and multi-model AI platforms
• Conceptual operation cost: Emphasizes irreversibility of permission changes

---

Reference Implementation

• Python + FastAPI + WebSocket prototype
• Supports device management, AI calls, upgrade requests, and audit logs
• Every gear shift records a conceptual operation cost

---

Contributing

• Fork the repository
• Create a feature branch (feature/your-feature)
• Submit a pull request
• Discuss major changes via Issues or Discussions
• See CONTRIBUTING.md for full guidelines

---

Roadmap

• v1.1: Audit log query API, customizable notifications
• v1.2: Industry compliance modules (Finance, Healthcare)
• v2.0: Automated compliance auditing, multi-model support

---

License

EEAL is licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0). You are free to share, copy, redistribute, remix, transform, and build upon the material for any purpose, even commercially, provided you give appropriate credit.

© 2026 EESCF

Full license text: https://creativecommons.org/licenses/by/4.0/legalcode