Skip to main content
Praxikon
Article 10 of 1139%
Nederlands

Article 10: Data and data governance

Under Article 10(3), the training, validation and testing data sets of a high-risk AI system must be relevant, sufficiently representative and, to the best extent possible, free of errors and complete in view of the intended purpose, with the appropriate statistical properties. Paragraph 4 requires them to reflect the specific geographical, contextual, behavioural or functional setting to the extent the intended purpose requires, and paragraph 2 calls for appropriate data governance and management practices. For systems that do not train AI models, paragraph 6 limits these requirements to the testing data sets.

, AI and privacy legal adviserobligations checked against the source on

Praxikon tracks Article 10 (Data and data governance) under the EU AI Act, citing the source for every statement.

Need help with high-risk AI? Get a signed assessment

EU Official:
UpcomingAnnex III from 2 Dec 2027; Annex I from 2 Aug 2028
Chapter III: High-risk AI systems

Application dates

  • : Requirements for high-risk systems listed in Annex III
  • : Requirements for high-risk systems linked to Annex I

Official text

||

Source: EUR-Lex, Regulation (EU) 2024/1689. Text reproduced verbatim.

Download AI Act (PDF) →

Official guidance on this article

2

Article 10 checklist

  • Training data is relevant and representative
  • Datasets checked for errors and bias
  • Data governance procedures established
  • Validation and test sets separated from training data
  • Data privacy protection ensured

Related recitals

Related tools

Related enforcement

No enforcement actions for this article yet. Follow developments via the Enforcement Tracker.

Related blog posts

Cross-references

Recitals

Annexes

Frequently asked questions

What requirements does the AI Act set for training data?+
Under Article 10(3), the training, validation and testing data sets of a high-risk AI system must be relevant, sufficiently representative and, to the best extent possible, free of errors and complete in view of the intended purpose, with the appropriate statistical properties. Paragraph 4 requires them to reflect the specific geographical, contextual, behavioural or functional setting to the extent the intended purpose requires, and paragraph 2 calls for appropriate data governance and management practices. For systems that do not train AI models, paragraph 6 limits these requirements to the testing data sets.
May I use special category personal data for bias detection in training data?+
Yes, exceptionally, but the legal basis is now Article 4a, not Article 10(5), which has been deleted. Under Article 4a(1), providers of high-risk AI systems may process special categories of personal data to the extent strictly necessary to detect and correct bias, subject to appropriate safeguards. Among other conditions, the bias detection and correction must not be effectively achievable with other data, including synthetic or anonymised data, and the data must be deleted once the bias is corrected or the retention period ends, whichever comes first.
Do I need to document the provenance of my training data?+
Yes, where relevant. Annex IV, point 2(d), requires the technical documentation under Article 11 to include, where relevant, information about the provenance, scope and main characteristics of the training data sets and how the data was obtained and selected, along with labelling procedures and data cleaning methodologies. Article 10(2), point (b), also calls for data governance practices covering data collection processes and the origin of data.
Who does Article 10 of the AI Act apply to?+
Article 10 sets requirements for the data sets used to develop high-risk AI systems and does not itself name a party. Under Article 16, point (a), providers must ensure their systems comply with the Section 2 requirements, which include Article 10. Article 10 imposes no obligation on deployers.
From when does Article 10 of the AI Act apply?+
Article 10 is in Chapter III, Section 2 of the AI Act (High-risk AI systems). Under Article 113, third paragraph, point (c), that section applies from 2 December 2027 for AI systems classified as high-risk pursuant to Article 6(2) and Annex III, and from 2 August 2028 for AI systems classified as high-risk pursuant to Article 6(1) and Annex I. Article 111(2) sets out the transitional rules for high-risk AI systems placed on the market or put into service before those dates, with 2 August 2030 as the final date for systems intended to be used by public authorities; Article 111(1) does so for the large-scale IT systems listed in Annex X.

What Article 10 requires in practice

Connections

What connects to Article 10 AI Act

Themes where this returns

Where this meets the other law

The counterpart in the other law

GDPR interpretation that also applies here6 of 12