Standards Framework
Published technical standards for responsible AI development and deployment
GCAIS develops and maintains technical standards for AI systems. Standards are developed through a multi-stage process involving technical review, public comment periods, and formal adoption by the Standards Board. Each standard undergoes review on a minimum 24-month cycle.
All standards documents are freely available. Organizations may assess their systems against published standards prior to submitting an accreditation application.
Development Methodology
Each GCAIS standard follows the same development process:
- Working Group Formation - Technical working group convened by the Standards Board
- Draft Publication - Draft standard published for a minimum 60-day public comment period
- Technical Review - Standards Board reviews comments and revises draft
- Adoption - Standards Board votes on adoption; two-thirds majority required
- Publication - Final standard published with version number and date stamp
- Review Scheduling - Review date set at adoption (minimum 24-month cycle)
Versioning Policy
GCAIS standards use semantic versioning. Major version increments (e.g., v1.x to v2.0) indicate substantive changes to requirements. Minor version increments (e.g., v1.0 to v1.1) indicate clarifications or non-substantive amendments. All versions are date-stamped at publication. Prior versions remain accessible for reference.
Current Standards
AI Safety Standard
GCAIS-STD-001 • v1.0 • 2026-01-15
Establishes minimum safety requirements for AI systems operating in production environments where outputs may influence human decisions or physical actions. Covers risk classification, failure mode documentation, output boundary testing, and incident reporting.
Read Standard →Transparency Standard
GCAIS-STD-002 • v1.1 • 2026-02-01
Requirements for disclosure of AI system capabilities, limitations, training data provenance, decision rationale, and user notification of AI involvement in processes.
Read Standard →Privacy & Data Handling
GCAIS-STD-003 • v1.0 • 2026-01-15
Standards for data collection minimization, retention limits, subject access rights, third-party data sharing controls, and cross-border transfer requirements.
Read Standard →Content Integrity
GCAIS-STD-004 • v1.0 • 2026-03-01
Requirements for AI systems that generate, modify, or distribute media content. Covers AI-origin labeling, provenance metadata, synthetic media watermarking, and content removal mechanisms.
Read Standard →Operational Quality
GCAIS-STD-005 • v1.0 • 2026-03-01
Performance reliability, uptime, error handling, incident management, and change management standards for production AI services. Covers SLA documentation and dependency management.
Read Standard →Human Oversight
GCAIS-STD-006 • v1.1 • 2026-02-20
Mandatory human oversight mechanisms for AI systems making or substantially influencing consequential decisions. Covers review triggers, override mechanisms, audit trails, and escalation paths.
Read Standard →