Skip to content
AI + Quantum Ethics Standards

Escalation, Pause, and Moratorium Mechanisms

Purpose

This document defines conditions under which development, deployment, or use of AI or quantum computing (QC) systems must be escalated, paused, or prohibited, with safeguards to prevent abuse.

Pauses are protective and evidence-driven, not punitive.

Ethical Mapping

  • A5 Proportionality & Moderation
  • A7 Stewardship
  • A3 Justice, Due Process, and Remedy
  • A6 Participation & Consultation

Definitions (Only as Required)

  • Escalation: a mandatory governance step requiring higher authority review before proceeding.
  • Pause: a temporary stop on a defined activity (development, deployment, or use) pending review and mitigation.
  • Moratorium: a pause of indefinite duration until explicit conditions for resumption are met, typically for Tier 3 risks.

Trigger Conditions

Trigger conditions are grouped into technical, ethical/legal, and societal/systemic categories. Any trigger MUST be documented with evidence and an initial risk-tier assessment.

Technical Triggers

T-1 (Critical Safety/Security Finding). A newly discovered vulnerability or failure mode that enables severe harm at scale, especially if exploitation is plausible and controls are insufficient.

T-2 (Loss of Control). Evidence that operators cannot reliably stop, roll back, or meaningfully constrain harmful outputs or actions in the deployed environment.

T-3 (Cryptographic Destabilization Signal). Evidence that QC capability or research materially increases feasibility of compromising widely deployed cryptographic protections for sensitive data with long confidentiality needs.

Ethical / Legal Triggers

E-1 (Rights-Violating Use). Credible evidence of use that violates human rights, due process, or meaningful consent in ways not remediable through immediate control changes.

E-2 (Irremediable Disparate Harm). Evidence of severe discriminatory impact that cannot be mitigated without major redesign or discontinuation.

Societal / Systemic Triggers

S-1 (Dual-Use Escalation). Capability diffusion or access changes that materially increase misuse potential beyond the originally reviewed threat model.

S-2 (Unacceptable Societal Disruption). Credible evidence of systemic destabilization risk (e.g., critical infrastructure impacts, large-scale coercion, irreversibility) exceeding Tier 3 tolerance.

Authority and Initiation

AP-1 (Initiation Rights). For Tier 2–3 systems, the following parties MUST be able to initiate escalation/pause:

  • the designated safety/security owner,
  • the system owner,
  • an independent audit function (where present),
  • a governance body defined in 01_governance/governance_model.md.

AP-2 (No Retaliation). Organizations MUST prohibit retaliation for good-faith escalation/pause initiation.

Escalation and Pause Process

P-1 (Immediate Containment). When a trigger condition is met, operators MUST implement immediate containment actions proportional to risk (e.g., disable feature, restrict access, revert version).

P-2 (Decision Record). Operators MUST create a decision record containing:

  • trigger evidence and classification
  • affected tiers and systems/versions
  • proposed mitigations and timeframe
  • residual risk statement
  • decision authority and sign-offs

P-3 (Consultation). For Tier 3 triggers, review MUST include multi-stakeholder input and, where relevant, independent security and human-rights review.

P-4 (Resumption Criteria). Pauses and moratoria MUST define resumption criteria and verification evidence. If criteria cannot be stated, the system SHOULD remain paused until criteria can be defined or the activity is discontinued.

Safeguards Against Abuse

SA-1 (Scope Boundaries). Pauses MUST be scoped to the minimum activity necessary to reduce harm, unless broader scope is justified by evidence.

SA-2 (Time-Bounded Review). Temporary pauses MUST include a review cadence and a target decision date.

SA-3 (Appeal/Contestability). Affected parties (including operators and external stakeholders where appropriate) SHOULD have a documented path to contest pause decisions, without undermining safety.

SA-4 (Transparency). Where feasible, operators SHOULD publish a summary of the pause decision and resumption criteria, with redactions limited to privacy/safety/legal constraints.

Compliance Evidence

  • trigger evidence and incident reports
  • containment actions and rollback logs
  • decision records and sign-offs
  • consultation artifacts (Tier 3)
  • public summaries/redaction logs (where published)