OECD Publishes Systemic Approach to Classifying AI Systems — June 8, 2022
The OECD released a methodology to classify AI systems by context, impact, and risk treatment, guiding regulators and developers on appropriate safeguards.
The working paper A Systemic Approach to the Classification of AI Systems introduced an adaptable taxonomy anchored in the OECD AI Principles. It maps AI applications across sectors, levels of autonomy, and potential adverse impacts to help policymakers calibrate oversight.
- Risk lenses. The framework evaluates AI systems through dimensions such as human agency, technical robustness, data governance, and societal impact.
- Policy tooling. Regulators can use the classification to align testing, certification, and incident reporting obligations with system characteristics.
- Standards alignment. The paper connects the taxonomy to emerging international standards and supports harmonised terminology across jurisdictions.
Product and compliance leads can apply the OECD taxonomy when mapping controls to AI Act, NIST AI RMF, and ISO/IEC 42001 requirements.