What happened
The OECD published a policy brief (published early November 2026, within the window) that "maps the principal mechanisms currently used to source data for training AI systems and proposes a taxonomy to support policy discussions." It sits inside the OECD/GPAI workstream on AI, data governance and privacy, building on the earlier "Enhancing user control and privacy protections in the context of AI" report and the OECD Expert Group on AI, Data, and Privacy. The taxonomy categorises the routes by which developers obtain training data (licensing, web scraping, and other collection mechanisms) and frames how each interacts with data quality, rights, and trustworthy-AI expectations under national AI governance frameworks.
Why it matters
Training-data provenance, copyright and privacy obligations are becoming board-level liabilities; this gives executives and policy leads a common OECD taxonomy for structuring licensing, scraping and data-access discussions.
Action needed
Map the OECD taxonomy against your model-provider and data-sourcing contracts to anticipate disclosure, licencing and rights obligations.