GlofAI 1000:2025

Responsible AI Standard (G-RAIS)

Framework for ethical AI adoption, fairness, bias reduction, and accountability – guidance for organisations to design, deploy and govern AI responsibly.

Status : Published Edition: 1.2025 Page : 48 Technical Committee: GlofAI/TC-1

What is G-RAIS?

G-RAIS (GlofAI 1000:2025) provides a practical, auditable framework for organisations to ensure AI systems are developed and used in ways that are fair, transparent, and accountable. It sets out principles, required governance practices and evidence expected for conformance.

Why responsible AI matters

Without consistent governance, AI systems can produce unfair outcomes, invade privacy, or create safety and reputational risks. G-RAIS helps organisations reduce those risks, meet regulator expectations, and build public trust.

Key benefits

Reduced bias & fairer outcomes
Clear governance & accountability
Better regulatory readiness and auditability
Increased stakeholder trust and vendor clarity

Who should use this standard?

G-RAIS is written for organisations of all sizes adopting AI — product teams, risk & compliance, legal, data scientists, procurement, and auditors. It is equally applicable to vendors, system integrators and procuring organisations.

Typical users

Enterprises deploying AI at scale
Public sector bodies procuring AI
Vendors wishing to demonstrate trustworthy AI

Use cases

AI model procurement & vendor selection
Internal governance & audit preparation
Regulatory impact assessment & reporting

Scope & structure

G-RAIS defines requirements and guidance across the AI lifecycle — from problem scoping and data governance through model development, deployment, monitoring and end-of-life. It is structured to be auditable and aligns with other GlofAI standards.

Core clauses

1. Introduction & scope (purpose, applicability)
2. Governance & accountability (roles, board oversight)
3. Ethical principles (fairness, explainability, human oversight)
4. Data & dataset management (quality, representativeness, lineage)
5. Model lifecycle & validation (testing, bias checks, metrics)
6. Security & privacy (data protection, access control)
7. Monitoring & incident response (drift detection, complaints)
8. Documentation & evidence (audit trail, reporting)

Implementation guidance

The standard includes pragmatic annexes and templates to help implementers accelerate adoption: a readiness checklist, a model-risk register template, sample data lineage table, and audit evidence examples.

Suggested implementation steps

1. Run a gap analysis using the G-RAIS readiness checklist.
2. Define governance roles and an AI governance board charter.
3. Integrate bias & fairness testing into CI/CD for models.
4. Establish monitoring dashboards and incident playbooks.
5. Conduct internal audits and prepare conformance evidence.

Frequently asked questions

Any organisation that develops, deploys or uses AI-based systems (including machine learning, generative AI, automated decision-making systems) and wishes to demonstrate that its AI systems comply with the G-RAIS framework’s requirements for fairness, transparency, accountability and governance can seek certification. The certification is typically carried out by a recognised auditing body or certification body that assesses the organisation’s practices, documentation, governance, risk-management processes and evidence of conformity to the standard.

G-RAIS is not inherently mandatory in the sense of being a legal requirement across all jurisdictions. It is a standard (framework) that organisations may choose to implement and certify against in order to demonstrate responsible AI practices, meet stakeholder expectations, or align with internal or regulatory governance goals. However, in some regulatory environments, compliance with frameworks like G-RAIS may be expected or strongly encouraged by regulators or industry bodies, so while the standard itself may not be legally mandatory everywhere, adopting it may become essential for meeting broader regulatory or market-requirements.

For models already in production, G-RAIS requires organisations to conduct a retrospective assessment: review and document the model’s development lifecycle, governance practices, risk-assessment, monitoring, performance, fairness/bias checks, transparency and accountability measures. The organisation needs to show how the model is being monitored in operation, how updates or re-training are managed, how risks are controlled (e.g., drift, bias, undesirable outcomes), and ensure that the production model and its operations meet the standard’s governance and evidence requirements. In effect, even existing models must be brought into alignment with the G-RAIS principles and documented accordingly to achieve certification.

Resources & downloads

G-RAIS Roadmap (One-Pager) checklist
G-RAIS self-audit guide pdf
Audit evidence collection template
Examples of use cases brief

Get help implementing G-RAIS

We run training, gap assessments and certification readiness workshops tailored to your organisation.

Quick Links

Purchase

Digital: free

PDF: $49.00

Buy Now Request Sample

Contact & support

Need implementation help or training?
Book a consultation.

Contact Advisors