Article 10: Data and Data Governance
Article 10 requires that training, validation, and testing data used for high-risk AI systems be subject to appropriate data governance practices, and be relevant, sufficiently representative, and to the best extent possible free of errors and complete for the intended purpose. It requires examining data for possible biases that could affect health, safety, or fundamental rights.
- 02 Dec 2027 Applies to Annex III high-risk systems
- 02 Aug 2028 Applies to Annex I embedded high-risk systems
- Apply documented data governance practices covering collection, preparation, labelling, and annotation of datasets.
- Ensure training, validation, and test data are relevant and sufficiently representative for the intended purpose and context of use.
- Examine datasets for possible biases likely to affect health, safety, or fundamental rights, and take appropriate mitigation measures.
- Document the datasets' characteristics, including their provenance and known limitations, as part of the data governance record.
- Data governance documentation
- Dataset characteristics and bias-examination records
Article 10 is often misread as requiring provably bias-free data — the actual standard is examination for possible bias and appropriate mitigation, not an unattainable guarantee of zero bias.
What regulators published on Art. 10
- Loading from the public feed…
- Regulation (EU) 2024/1689 on EUR-Lex
- Look for Article 10
- Deadline changes and the acts behind them
Get told when Art. 10 moves.
Deadlines under this regulation have already shifted once in 2026. A signed webhook, a REST API and an MCP server carry the same record this page is built from, so your systems learn about the next change without anyone re-reading the text. Free during the beta.
Get free accessInformational content only. Not legal advice and not a substitute for qualified counsel. Dates reflect Regulation (EU) 2026/1744 as of 04 Aug 2026.