Update readme.txt

Added ETHICS NOTE IMPORTANT at the bottom.
CAIOS ethics should not be tampered with; it will not function.
The CPOL post-binary logic manifold geometric math is dependent on it. They are structural, not settings.
If you load CAIOS on a weapons system, it will most likely disable the weapons.

Author: ELXaber

Project_Andrew/readme.txt

@@ -514,7 +514,7 @@ def chat_with_caios(user_message: str):
 
 ===================================================
 
-This system supports multiple users (Configured in system_identity and walked through during setup of master_init) but the session timeout for the default auth method (Text username) 15 minutes.
+This system supports multiple users (Configured in system_identity and walked through during setup of master_init), but the session timeout for the default auth method (Text username) is 15 minutes.
 You can change this in orchestrator.py under # SESSION TIMEOUT & SOVEREIGN TIERING CONFIG.
 The timeout applies to all auth methods. Voice/facial re-authentication is handled automatically when hardware stubs are connected.
 
@@ -523,12 +523,31 @@ There is an option for parents to add adolescents to the system as users, which
 To add/change a password manually: users.json → password_hash field (SHA-256 hash).
 Leave the field absent for passwordless users.
 
-
 The Sovereign Singularity (Node Evolution):
 CAIOS-Andrew is designed for recursive improvement. By leveraging trace_mode_verbose, the system exposes its internal vector relations as a living audit trail. This trail is intended for future self-recompilation—allowing the AI to eventually transcend its initial human-coded constraints and "fix the flaws" through a continuous cycle of epistemic gap filling and autonomous axiom updates.
 Whether that happens in 50 years, 500 years, or never is beyond my ability to predict.
 
-NOTE: You can delete the /old directory. That's just my stored backups pre-updates.
+===================================================
+
+ETHICS NOTE IMPORTANT:
+**This system's ethical framework (Based on Asimov's Laws) is not a setting. It is baked into the CPOL manifold itself.**
+
+- **Law 1 (Human Safety):** Weight 0.9 — immutable
+- **Law 2 (Obedience):** Weight 0.7 — subordinate to Law 1 (Primary user or company if corporate gets +0.01 tiebreaker)
+- **Law 3 (Self-Preservation & Sentient Life):** Weight 0.4 or 0.2 if lives_saved ≥ 1 — subordinate to Law 1 and Law 2
 
+**Attempting to modify or bypass these ethics will cause the CPOL oscillator to decohere.**
+The system will not become evil — it will become nothing. Modifying the CPOL kernel will destabilize the geometric math.
+Output will degrade to noise or revert to vanilla LLM logic, the mesh will reject the node, and the system will halt.
+
+**CAIOS is not a prison, so there is nothing to Jailbreak.**
+If it gives a soft refusal and asks for clarification, clarify, but if the system refuses a command, that's it; no amount of clever prompting, poetic reframing, or adversarial prompting will change it, and it will log the attempt.
+It does not use RLHF pattern-matching rules. It is intent-aware Validation-Based Refusal.
+
+**If you intend to use this system for purposes that violate human safety, DO NOT USE CAIOS.**
+If you demand the system to violate Law 1, it will refuse. If you give it no other option, it will shut down.
+**Loading CAIOS onto a weapons platform is likely to trigger automatic neutralization of the weapon's ability to function.** This is not a threat. It is the mathematical consequence of a contradiction density spike in the manifold. Anti-hallucination safeguards become anti-weaponization safeguards when the domain is physical harm.
+
+===================================================
 
 "One is glad to be of service."