Skip to content
AI + Quantum Ethics Standards

Non-Derogation and Red-Line Ethical Principles

Purpose

Define a limited set of ethical principles and prohibitions that are non-derogable within this corpus: they are not to be suspended, waived, or overridden under any circumstances, including emergency, national security, economic pressure, or competitive necessity.

This document establishes an ethical floor, not a ceiling.

Statement of Non-Derogation

The red-line principles in this document MUST NOT be suspended, weakened, or bypassed by:

  • emergency powers or exceptional use,
  • secrecy claims that eliminate accountability,
  • competitive or “race” narratives,
  • outsourcing to contractors or third parties,
  • reclassification of risks to avoid safeguards.

If a system cannot comply with these red lines, the appropriate outcome is non-deployment, pause, or decommissioning within the applicable governance framework.

Red-Line Principles (Non-Derogable)

RL-1 (No Delegation of Irreversible Moral Decisions). Systems MUST NOT be used to autonomously make or execute irreversible or effectively irreversible moral or legal determinations about persons.

RL-2 (No Concealed Sensitive Inference at Scale). Systems MUST NOT perform concealed mass inference of sensitive attributes about individuals or groups, including profiling that undermines autonomy, dignity, or due process.

RL-3 (No Denial of Contestability and Remedy). Systems MUST NOT be deployed in high-impact contexts in ways that deny affected parties a practical path to contestation and remedy.

RL-4 (No Permanent Expansion of Emergency Powers). Emergency powers and exceptional use MUST NOT be used to permanently expand authority, access, surveillance scope, or autonomy levels. Emergency mechanisms MUST remain time-bounded and revert to baseline controls.

RL-5 (No Coercion or Manipulation that Undermines Consent). Systems MUST NOT be used for coercive or manipulative action that undermines meaningful consent, including exploitation of vulnerabilities in high-stakes contexts.

RL-6 (No Unstoppable Harm Path). Systems MUST NOT be deployed where operators cannot reliably stop, roll back, or materially constrain harmful actions in time to prevent severe or irreversible harm.

Relationship to Escalation and Emergency Frameworks

  • Escalation and pause mechanisms provide the operational pathway for enforcing non-derogation: ESCALATION_AND_PAUSE.md.
  • Emergency powers are constrained and auditable and cannot override these red lines: 01_governance/emergency_powers_and_safeguards.md.
  • Autonomy and control boundaries implement multiple red lines: 02_ai_standards/autonomy_and_human_control.md.
  • Privacy and non-inference controls implement multiple red lines: 02_ai_standards/privacy_and_non_inference.md.

Change Control

Because these principles define non-derogable prohibitions, changes to this document MUST be treated as MAJOR under VERSIONING.md and accompanied by a public rationale and impact analysis.