The Model AI Governance Framework for Generative AI is a non-binding governance framework issued on 30 May 2024 by Singapore's Infocomm Media Development Authority (IMDA) and the AI Verify Foundation. It extends the 2019/2020 Model AI Governance Framework to cover generative and foundation models and is structured around nine dimensions. The document is presented as non-prescriptive guidance designed to interoperate with other governance regimes rather than as binding law.
Summary of the framework
The framework identifies nine dimensions for a trusted generative-AI ecosystem:
- Accountability — a shared-responsibility structure across model developers, application deployers, and cloud providers, drawn by analogy from cloud computing's Shared Responsibility Model.
- Data — data quality, trusted sources, respect for personal data and copyright, and dataset governance.
- Trusted Development and Deployment — evaluation, safety fine-tuning, system and model cards, and lifecycle disclosure.
- Incident Reporting — structured processes modelled on aviation and cybersecurity incident regimes.
- Testing and Assurance — third-party testing and common standards, complementing the AI Verify toolkit.
- Security — prompt injection, model theft, adversarial attacks, and misuse vectors.
- Content Provenance — watermarking, C2PA-style cryptographic provenance, and labelling.
- Safety and Alignment R&D — accelerated through global cooperation among AI Safety Institutes.
- AI for Public Good — democratised access, public-sector adoption, worker upskilling, and sustainability.
The framework was developed through a public consultation running from 16 January to 15 March 2024 and through engagement with the Frontier Model Forum and industry. Companion initiatives include the AI Verify testing framework (2023) and the Gen AI Evaluation Sandbox.
Key claims
The framework positions Singapore to interoperate with existing governance regimes rather than to author a distinct rulebook: it is explicitly designed to align with the OECD AI Principles, the G7 Hiroshima Code, the NIST AI RMF, and the EU AI Act and GPAI Code of Practice.
The accountability dimension allocates roles across the model developer, the application deployer, and the cloud provider, an approach the framework adapts from cloud-governance doctrine's Shared Responsibility Model.
Incident reporting is treated as a core element, modelled on aviation and cybersecurity incident regimes. This approach parallels US state-level incident-reporting measures such as California SB 53 — Transparency in Frontier AI Act.
Dimension 8 references global cooperation among AI Safety Institutes, situating the framework's safety and alignment goals within AISI-led infrastructure.
The framework is non-prescriptive guidance rather than binding law, consistent with Singapore's broader treatment of AI as a growth strategy.
Relation to other frameworks
The document is the first Singapore-authored governance framework summarized here, alongside US (NIST AI Risk Management Framework (AI RMF 1.0)), EU (EU AI Act (Regulation 2024/1689)), and UK (The Bletchley Declaration (AI Safety Summit, 1–2 November 2023)) instruments. Its nine-dimension structure offers a point of comparison against the 11-action G7 Hiroshima Code and the 8-commitment Seoul Commitments. The content-provenance dimension (dimension 7) aligns with action 7 of the G7 Hiroshima Code of Conduct for Advanced AI (2023) and related watermarking discussions, the incident-reporting dimension (dimension 4) connects to California SB 53 — Transparency in Frontier AI Act and Managing Advanced Cyber Risks in Frontier AI Frameworks, and dimension 8 connects to AI Safety Cases and Frameworks.
Provenance
Issued 30 May 2024 by IMDA and the AI Verify Foundation, following the 16 January – 15 March 2024 public consultation. The framework page is Singapore Model AI Governance Framework for Generative AI (2024).
Relationships
- supports: Singapore Model AI Governance Framework for Generative AI (2024) (framework page), AI Safety Cases and Frameworks
- depends-on: G7 Hiroshima Code of Conduct for Advanced AI (2023), NIST AI Risk Management Framework (AI RMF 1.0)
- related: EU AI Act (Regulation 2024/1689), EU General-Purpose AI Code of Practice (Final Version, 2025), The Bletchley Declaration (AI Safety Summit, 1–2 November 2023), California SB 53 — Transparency in Frontier AI Act, Managing Advanced Cyber Risks in Frontier AI Frameworks, Infocomm Media Development Authority (IMDA, Singapore)