Skip to content

Regulation Tracker / EU AI Act / Chapter III · Section 2

Article 10
Data and data governance

Chapter III — Classification rules for high-risk AI systems, Section 2 — Compliance with the requirements. 9 distinct duties, powers or definitions are coded from this article, applying from 2027-12-02. 9 reach legal practice directly.

Amended by Regulation (EU) 2026/1744 (Digital Omnibus on AI), in force 2026-07-27 — about 314 words changed. The text below is the consolidated version of 27 July 2026; the 2024 text and the amending regulation are on EUR-Lex. 9 of the duties below now apply later than Article 113 originally provided; each shows both dates.

official text: EUR-Lex (consolidated 27 Jul 2026) · text © European Union (Decision 2011/833/EU) · coding CC BY 4.0 · data: obligations.json · Hugging Face · GitHub

The text

1. High-risk AI systems which make use of techniques involving the training of AI models with data shall be developed on the basis of training, validation and testing data sets that meet the quality criteria referred to in paragraphs 2, 3 and 4 of this Article and in Article 4a(1) whenever such data sets are used.

2. Training, validation and testing data sets shall be subject to data governance and management practices appropriate for the intended purpose of the high-risk AI system. Those practices shall concern in particular:

  1. (a)the relevant design choices;
  2. (b)data collection processes and the origin of data, and in the case of personal data, the original purpose of the data collection;
  3. (c)relevant data-preparation processing operations, such as annotation, labelling, cleaning, updating, enrichment and aggregation;
  4. (d)the formulation of assumptions, in particular with respect to the information that the data are supposed to measure and represent;
  5. (e)an assessment of the availability, quantity and suitability of the data sets that are needed;
  6. (f)examination in view of possible biases that are likely to affect the health and safety of persons, have a negative impact on fundamental rights or lead to discrimination prohibited under Union law, especially where data outputs influence inputs for future operations;
  7. (g)appropriate measures to detect, prevent and mitigate possible biases identified according to point (f);
  8. (h)the identification of relevant data gaps or shortcomings that prevent compliance with this Regulation, and how those gaps and shortcomings can be addressed.

3. Training, validation and testing data sets shall be relevant, sufficiently representative, and to the best extent possible, free of errors and complete in view of the intended purpose. They shall have the appropriate statistical properties, including, where applicable, as regards the persons or groups of persons in relation to whom the high-risk AI system is intended to be used. Those characteristics of the data sets may be met at the level of individual data sets or at the level of a combination thereof.

4. Data sets shall take into account, to the extent required by the intended purpose, the characteristics or elements that are particular to the specific geographical, contextual, behavioural or functional setting within which the high-risk AI system is intended to be used.

6. For the development of high-risk AI systems not using techniques involving the training of AI models, paragraphs 2, 3 and 4 of this Article and Article 4a(1) shall apply only to the testing data sets.

What it requires, coded

SafeLegalAI's reading of each duty in this article: who, what, from when, under which fine tier, and whether it reaches a firm, chambers, court or legal-AI vendor. Descriptive, not advice; the quoted words are the Regulation's.

  1. Art. 10(1)Data governanceHigh-risklegal practice

    High-risk AI systems using training techniques must be developed using training, validation and testing datasets that meet Article 10(2), (3) and (4) quality criteria and Article 4a(1) when used.

    "High-risk AI systems which make use of techniques involving the training of AI models with data shall be developed on the basis of training, validation and testing data sets that meet the quality criteria referred to in paragraphs 2, 3 and 4 of this Article and in Article 4a(1) whenever such data sets are used."
    Who
    Provider
    From
    (Art. 113(c)(i) as amended by Reg. (EU) 2026/1744)As enacted in 2024: (Art. 113 second subparagraph); deferred by Regulation (EU) 2026/1744 (Digital Omnibus on AI).For Art. 6(1)/Annex I high-risk systems, Art. 113(c)(ii) as amended applies the corresponding obligation from 2028-08-02; Art. 111(2) gives public-authority high-risk systems until 2030-08-02.
    Fine tier
    Art. 99(4)
    Legal practice
    Reaches legal practice directly — Key for legal-AI vendors training high-risk judicial, ADR or public-sector systems on legal, case or user data.
    See
    Art. 4a(1) · Art. 10(2) · Art. 10(3) · Art. 10(4) · Annex III point 8

  2. Art. 10(2)Data governanceHigh-risklegal practice

    Training, validation and testing datasets must be subject to governance and management practices appropriate for the intended purpose of the high-risk AI system.

    "Training, validation and testing data sets shall be subject to data governance and management practices appropriate for the intended purpose of the high-risk AI system. Those practices shall concern in particular:"
    Who
    Provider
    From
    (Art. 113(c)(i) as amended by Reg. (EU) 2026/1744)As enacted in 2024: (Art. 113 second subparagraph); deferred by Regulation (EU) 2026/1744 (Digital Omnibus on AI).For Art. 6(1)/Annex I high-risk systems, Art. 113(c)(ii) as amended applies the corresponding obligation from 2028-08-02; Art. 111(2) gives public-authority high-risk systems until 2030-08-02.
    Fine tier
    Art. 99(4)
    Legal practice
    Reaches legal practice directly — Data governance is central for legal-AI vendors using case law, pleadings, client files or public records in high-risk systems.
    See
    Annex III point 8

  3. Art. 10(2) (a-e)Data governanceHigh-risklegal practice

    Dataset governance must address design choices, data collection and origin, data preparation, assumptions about what data represent, and availability, quantity and suitability of needed datasets.

    "the relevant design choices;"
    Who
    Provider
    From
    (Art. 113(c)(i) as amended by Reg. (EU) 2026/1744)As enacted in 2024: (Art. 113 second subparagraph); deferred by Regulation (EU) 2026/1744 (Digital Omnibus on AI).For Art. 6(1)/Annex I high-risk systems, Art. 113(c)(ii) as amended applies the corresponding obligation from 2028-08-02; Art. 111(2) gives public-authority high-risk systems until 2030-08-02.
    Fine tier
    Art. 99(4)
    Legal practice
    Reaches legal practice directly — Legal-AI vendors need traceable data lifecycle governance for high-risk systems trained or tested on legal-domain data.
    See
    Art. 10(2)(a)-(e) · Annex III point 8

  4. Art. 10(2) (f)Data governanceHigh-risklegal practice

    Providers must examine datasets for biases likely to affect health and safety, negatively affect fundamental rights or lead to discrimination prohibited under Union law.

    "examination in view of possible biases that are likely to affect the health and safety of persons, have a negative impact on fundamental rights or lead to discrimination prohibited under Union law, especially where data outputs influence inputs for future operations;"
    Who
    Provider
    From
    (Art. 113(c)(i) as amended by Reg. (EU) 2026/1744)As enacted in 2024: (Art. 113 second subparagraph); deferred by Regulation (EU) 2026/1744 (Digital Omnibus on AI).For Art. 6(1)/Annex I high-risk systems, Art. 113(c)(ii) as amended applies the corresponding obligation from 2028-08-02; Art. 111(2) gives public-authority high-risk systems until 2030-08-02.
    Fine tier
    Art. 99(4)
    Legal practice
    Reaches legal practice directly — Highly relevant to legal-AI vendors because biased legal-data outputs can influence future decisions about litigants, clients, workers or public-service users.
    See
    Art. 4a · Annex III point 8

  5. Art. 10(2) (g)Data governanceHigh-risklegal practice

    Providers must take appropriate measures to detect, prevent and mitigate possible biases identified under Article 10(2)(f).

    "appropriate measures to detect, prevent and mitigate possible biases identified according to point (f);"
    Who
    Provider
    From
    (Art. 113(c)(i) as amended by Reg. (EU) 2026/1744)As enacted in 2024: (Art. 113 second subparagraph); deferred by Regulation (EU) 2026/1744 (Digital Omnibus on AI).For Art. 6(1)/Annex I high-risk systems, Art. 113(c)(ii) as amended applies the corresponding obligation from 2028-08-02; Art. 111(2) gives public-authority high-risk systems until 2030-08-02.
    Fine tier
    Art. 99(4)
    Legal practice
    Reaches legal practice directly — High-risk legal-AI providers must address dataset bias that could affect equal treatment, due process or access to services.
    See
    Art. 10(2)(f) · Art. 4a · Annex III point 8

  6. Art. 10(2) (h)Data governanceHigh-risklegal practice

    Dataset governance must identify relevant data gaps or shortcomings that prevent compliance and how those gaps and shortcomings can be addressed.

    "the identification of relevant data gaps or shortcomings that prevent compliance with this Regulation, and how those gaps and shortcomings can be addressed."
    Who
    Provider
    From
    (Art. 113(c)(i) as amended by Reg. (EU) 2026/1744)As enacted in 2024: (Art. 113 second subparagraph); deferred by Regulation (EU) 2026/1744 (Digital Omnibus on AI).For Art. 6(1)/Annex I high-risk systems, Art. 113(c)(ii) as amended applies the corresponding obligation from 2028-08-02; Art. 111(2) gives public-authority high-risk systems until 2030-08-02.
    Fine tier
    Art. 99(4)
    Legal practice
    Reaches legal practice directly — Legal-domain datasets often have coverage, jurisdictional, demographic or quality gaps relevant to high-risk legal-AI compliance.
    See
    Annex III point 8

  7. Art. 10(3)Data governanceHigh-risklegal practice

    Training, validation and testing datasets must be relevant, sufficiently representative, as error-free and complete as possible, and have appropriate statistical properties for intended use.

    "Training, validation and testing data sets shall be relevant, sufficiently representative, and to the best extent possible, free of errors and complete in view of the intended purpose. They shall have the appropriate statistical properties, including, where applicable, as regards the persons or groups of persons in relation to whom the high-risk AI system is intended to be used. Those"
    Who
    Provider
    From
    (Art. 113(c)(i) as amended by Reg. (EU) 2026/1744)As enacted in 2024: (Art. 113 second subparagraph); deferred by Regulation (EU) 2026/1744 (Digital Omnibus on AI).For Art. 6(1)/Annex I high-risk systems, Art. 113(c)(ii) as amended applies the corresponding obligation from 2028-08-02; Art. 111(2) gives public-authority high-risk systems until 2030-08-02.
    Fine tier
    Art. 99(4)
    Legal practice
    Reaches legal practice directly — Legal-AI vendors must account for representativeness and quality when datasets affect courts, ADR, employment or public-service decisions.
    See
    Annex III point 8

  8. Art. 10(4)Data governanceHigh-risklegal practice

    Datasets must take into account, to the extent required by intended purpose, characteristics particular to the geographical, contextual, behavioural or functional setting of use.

    "Data sets shall take into account, to the extent required by the intended purpose, the characteristics or elements that are particular to the specific geographical, contextual, behavioural or functional setting within which the high-risk AI system is intended to be used."
    Who
    Provider
    From
    (Art. 113(c)(i) as amended by Reg. (EU) 2026/1744)As enacted in 2024: (Art. 113 second subparagraph); deferred by Regulation (EU) 2026/1744 (Digital Omnibus on AI).For Art. 6(1)/Annex I high-risk systems, Art. 113(c)(ii) as amended applies the corresponding obligation from 2028-08-02; Art. 111(2) gives public-authority high-risk systems until 2030-08-02.
    Fine tier
    Art. 99(4)
    Legal practice
    Reaches legal practice directly — Legal systems are jurisdictional and context-sensitive, making this important for legal-AI vendors deploying across courts, languages or practice areas.
    See
    Annex III point 8

  9. Art. 10(6)Data governanceHigh-risklegal practice

    For high-risk AI systems developed without model-training techniques, Article 10(2), (3) and (4) and Article 4a(1) apply only to testing datasets.

    "For the development of high-risk AI systems not using techniques involving the training of AI models, paragraphs 2, 3 and 4 of this Article and Article 4a(1) shall apply only to the testing data sets."
    Who
    Provider
    From
    (Art. 113(c)(i) as amended by Reg. (EU) 2026/1744)As enacted in 2024: (Art. 113 second subparagraph); deferred by Regulation (EU) 2026/1744 (Digital Omnibus on AI).For Art. 6(1)/Annex I high-risk systems, Art. 113(c)(ii) as amended applies the corresponding obligation from 2028-08-02; Art. 111(2) gives public-authority high-risk systems until 2030-08-02.
    Fine tier
    Art. 99(4)
    Legal practice
    Reaches legal practice directly — Relevant to rule-based or retrieval-heavy legal-AI systems where testing data, rather than training data, carries the Article 10 governance burden.
    See
    Art. 4a(1) · Art. 10(2) · Art. 10(3) · Art. 10(4) · Annex III point 8

Cite

Regulation (EU) 2024/1689, Article 10 (Data and data governance), as amended by Regulation (EU) 2026/1744, consolidated text of 27 July 2026, https://eur-lex.europa.eu/eli/reg/2024/1689/2026-07-27/eng#art_10 — text © European Union; only the Official Journal is authentic. Coding: SafeLegalAI (published by Cognesio LLP), "EU AI Act, structured", safelegalai.com/regulation/eu-ai-act/article-10, accessed 2026-09-08, CC BY 4.0.