Frameworks
AI governance frameworks and principles
Soft-law instruments, process frameworks and standards that guide AI governance. None of these entries should be read as legislation unless a page explicitly says the instrument is legally binding.
15 entries
- UNESCO Recommendation on the Ethics of Artificial IntelligenceInternational recommendation (soft law) · 2021 — Adopted by UNESCO’s General Conference in November 2021, the Recommendation is a comprehensive ethics instrument covering values and principles (including human rights and human dignity, living in harmony, diversity, and environment) and policy action areas such as data, health, culture, education, economy, and communication. It is a recommendation, not a treaty: it guides states but does not by itself create court-enforceable international obligations of the kind a convention would.
- OECD AI PrinciplesIntergovernmental principles / OECD Recommendation · 2019 — The OECD Recommendation on Artificial Intelligence sets principles for responsible stewardship of trustworthy AI (inclusive growth, human-centred values, transparency, robustness, and accountability) and national-policy recommendations on investment, ecosystems, enabling policy, capacity, and international co-operation. Adherents include OECD members and other countries that have adhered. The Principles are a leading reference for national AI strategies; they are not, by themselves, domestic criminal or civil law.
- NIST AI Risk Management FrameworkVoluntary risk-management framework · 2023 — The NIST AI Risk Management Framework (AI RMF 1.0) offers a process model—Govern, Map, Measure, Manage—to help organizations address AI risks. It is explicitly voluntary and not a U.S. federal regulation. Companion profiles, including for generative AI, tailor the framework to contexts. Organizations often use it alongside legal duties that come from other sources (sectoral law, contracts, or foreign regulation).
- G7 Hiroshima AI ProcessIntergovernmental process and codes of conduct · 2023 — Launched under Japan’s 2023 G7 presidency, the Hiroshima AI Process develops international guiding principles and a voluntary code of conduct for organizations developing advanced AI systems. It is diplomatic and voluntary. It should not be described as a global AI law. Outcomes are intended to complement domestic regulation and multi-stakeholder standards.
- Global Digital CompactIntergovernmental political compact · 2024 — The Global Digital Compact, agreed in the Pact for the Future process in 2024, sets shared objectives on digital cooperation, including AI, connectivity, data and human rights online. It is a political agreement among UN Member States, not a legally binding convention. Implementation runs through UN and national processes, including follow-up on AI governance dialogues.
- UN Global Dialogue on AI GovernanceInternational dialogue / institutional process · 2024 — The Global Dialogue on AI Governance is a United Nations platform for inclusive discussion of international AI governance, bringing together governments and other stakeholders. It is a process and convening mechanism, not a regulator and not a substitute for treaties or domestic law. Official materials and meeting outcomes should be read as diplomacy and agenda-setting.
- Ethics Guidelines for Trustworthy AIExpert ethics guidelines · 2019 — These guidelines set out that trustworthy AI should be lawful, ethical and robust, and list requirements including human agency, privacy, transparency, diversity, environmental well-being and accountability. They influenced later EU policy language but are not the AI Act. Treating the guidelines as if they were regulation misstates their legal status.
- ISO/IEC 42001 — AI management systemsInternational management-system standard · 2023 — ISO/IEC 42001 specifies requirements for establishing, implementing, maintaining and continually improving an AI management system. Certification, where offered, attests to the management system—not to the harmlessness of every model. The standard is a tool for organizational governance and can support procurement or regulatory conformity narratives when those regimes point to standards.
- Singapore Model AI Governance FrameworkNational voluntary model framework · 2020 — Singapore’s Model AI Governance Framework provides practical guidance on deploying AI in a responsible manner, organized around principles such as transparency, fairness and accountability, with implementation and communication advice for organizations. It is a national soft-law / guidance product, not Singapore criminal law and not extra-territorial legislation.
- UN General Assembly resolution on seizing the opportunities of safe, secure and trustworthy AIGeneral Assembly resolution · 2024 — In March 2024 the UN General Assembly adopted a resolution on safe, secure and trustworthy artificial intelligence. General Assembly resolutions express political agreement among Member States. They are important for agenda-setting and diplomatic alignment but are not, standing alone, binding treaties. Readers should consult the official resolution text for operative paragraphs.
- General-Purpose AI Code of PracticeCode of practice (voluntary implementation under the AI Act) · 2025 — The Code of Practice is intended to help providers of general-purpose AI models demonstrate how they meet relevant obligations under the EU AI Act. It is implementation guidance and a code, not a replacement for Regulation (EU) 2024/1689. Read the Commission page and the Act together.
- AI PactVoluntary implementation initiative · 2024 — The AI Pact is a Commission initiative for organizations to prepare for the AI Act ahead of staged application dates. It is not Union law and does not create the same obligations as the regulation. Treat it as an implementation pathway, not as a substitute for the Official Journal text.
- NIST Generative AI Profile (AI RMF)Voluntary risk-management profile · 2024 — NIST AI 600-1 (Generative Artificial Intelligence Profile) tailors the AI Risk Management Framework to generative-AI risks. It is guidance from a standards body, not a U.S. statute. Use it with the core AI RMF, not as a stand-alone legal duty.
- GAO Artificial Intelligence Accountability Framework (GAO-21-519SP)Accountability / audit framework · 2021 — GAO-21-519SP sets an accountability-oriented framework for federal agencies and other entities that use AI. It is an audit and oversight product, useful for public-sector governance, and is not itself binding law.
- A Pro-Innovation Approach to AI Regulation (UK)National policy white paper · 2023 — The UK white paper sets a principles-based, cross-sector approach in which existing regulators apply AI principles rather than creating a single horizontal AI statute in the EU AI Act sense. Later official guidance for regulators implements that approach. It is policy, not primary legislation.
Have an official resource, framework, law or other source to propose for a section of this directory? Use the suggestion form (opens in a new tab). Inclusion is editorial and not guaranteed.