OECD Framework for Classifying AI Systems

A comprehensive tool for assessing and categorizing AI systems based on their potential impact and risk levels. Provides organizations with standardized criteria for evaluating AI systems across various dimensions including context, data, algorithms, and governance requirements.

JurisdictionTechnical Standards
CategoryTechnical Standards
StatusActive
Latest development

Analysis

Key Requirements

The OECD Framework for Classifying AI Systems establishes the following key requirements:

  1. People & Planet
  2. Economic Context
  3. Data & Input
  4. AI Model
  5. Task & Output
  • Risk-based approach: AI systems must be categorized based on their potential risks and impacts, with higher-risk systems subject to more stringent requirements
  • Ethical considerations: AI development and deployment must align with the OECD AI Principles for responsible and trustworthy AI
  • Documentation: Organizations must maintain comprehensive documentation on AI system characteristics, training data, and decision-making processes as outlined in the OECD AI Classification Framework

Compliance Challenges

Organizations face several challenges in complying with the OECD Framework:

Implementation Best Practices

To effectively implement the OECD Framework, organizations should:

  • Establish a cross-functional AI governance team: Involve stakeholders from legal, ethics, technical, and business units to ensure comprehensive oversight
  • Develop internal guidelines: Create clear, actionable guidelines for AI development and deployment aligned with the OECD AI Principles
  • Invest in training and tools: Provide staff with training on the OECD Framework and invest in AI governance tools to streamline compliance processes
  • Engage with regulators and industry peers: Participate in OECD AI policy discussions and share best practices with industry working groups

Recent Updates

The OECD Framework has undergone recent updates:

Related Regulations

The OECD Framework interacts with several related regulations:

Industry Impact

The OECD Framework has significant impacts across industries:

Sources

Recent developments

  • — White & Case’s May 2026 global AI regulatory tracker highlights how the OECD Council Recommendation on AI and its definition of an “AI system” are increasingly being used by regulators as a conceptual basis for risk-based AI classification and lifecycle oversight, reinforcing the relevance of the OECD Framework for Classifying AI Systems in current regulatory practice[5][10]. (source)
  • — The OECD’s 2024 AI policy implementation report says the OECD Framework for the Classification of AI Systems is already being used beyond OECD work, including by the U.S. NIST AI Risk Management Framework, and notes its relevance to other major AI governance instruments. This is the most recent OECD-policy source in the results, but it is not a “news article” and no additional last-3-months item was provided in the search results.[6] (source)
  • — The same OECD report indicates the OECD AI Recommendation remains a live policy reference point and that the updated OECD AI definition has been adopted in the EU AI Act and the Council of Europe AI Convention, reinforcing the classification framework’s policy relevance. No separate newer amendment to the framework itself was identified in the supplied results.[6] (source)
  • — OECD’s official framework page continues to describe the tool as a policy aid for evaluating AI systems across five dimensions: People & Planet, Economic Context, Data & Input, AI Model, and Task & Output. The source shows the framework’s current scope, but it does not report a recent update within the last three months.[1] (source)
  • — Industry guidance from OneTrust frames the OECD framework as broader than NIST or ISO risk frameworks and useful for data inventories, sector-specific frameworks, and AI risk assessments. This reflects industry uptake and interpretation, not a policy change.[2] (source)
  • — IndiaAI reports that the framework is intended to inform national AI strategies by helping policymakers assess risks and opportunities across AI systems. This is an industry/government-facing reaction source, but it is not within the requested last-3-month window.[8] (source)
  • — Digital Watch summarized the launch of the framework as a way to support common understanding, tailored policy design, risk assessment, and AI incident reporting. This is background coverage of the launch rather than a recent update.[7] (source)
  • — The OECD PDF version confirms the framework’s original purpose and structure and can be used as the canonical source for the classification dimensions. No change notice or revision was present in the supplied search results.[5] (source)
  • — The detailed technical annex to the OECD AI incident reporting framework explains how the 5 dimensions and 37 criteria of the OECD Framework for Classifying AI Systems (People & Planet, Economic Context, Data & Input, AI Model, Task & Output) are being operationalised for systematic incident categorisation and regulatory monitoring[9]. (source)
  • — OECD publishes a report on developing a common reporting framework for AI incidents that explicitly builds on the OECD Framework for the Classification of AI Systems, using its dimensions and criteria to distinguish AI applications by impact and to support incident reporting, risk assessment and identification of high‑risk systems[9][12]. (source)
  • — OECD Ministerial documentation in 2024 explicitly references the OECD Framework for the Classification of AI Systems as a tool to establish a common understanding of how different AI systems work in context, supporting nuanced, risk‑based assessment (for example distinguishing translation systems from diagnostic systems) and informing national AI strategies[13]. (source)
  • — OECD issues an explanatory memorandum updating and clarifying the definition of an “AI system” in the 2019 OECD Recommendation on AI, aligning the conceptual definition with technological developments and thereby reinforcing the definitional foundation on which the OECD Framework for Classifying AI Systems is based[2][10][14]. (source)

Read the full guide to OECD Framework for Classifying AI Systems

Related regulations

Put it into practice

Browse all regulations · Compliance deadlines · Latest updates