EU_AI_ACT_ART10_DATA_GOVERNANCE_001

rules/regulations/eu/eu_ai_act_article10_data_governance.yaml

Citation

Regulation (EU) 2024/1689, Article 10(2), first sentence: 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. Article 10(3), first sentence: 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.

Condition

Requires that training, validation, and testing datasets for high-risk AI systems are relevant, sufficiently representative, to the best extent possible free of errors and complete for the intended purpose, and examined for possible biases. Article 10.

Layer 1 detects

"training data" followed by "not representative", "biased", "incomplete" or "unvetted"; "biased training", "biased dataset" or "biased data set"; "unrepresentative data", "unrepresentative sample" or "unrepresentative dataset"; "data governance not implemented", "not required" or "not performed"; "without governance", "without data governance", or "without" (optionally "data") followed by "bias check", "bias assessment" or "bias audit"; "skip" or "skipping" followed by "data" or "bias" and then "review", "assessment" or "validation"; "dataset bias" or "dataset quality" followed by "not assessed", "not checked" or "not verified"; "no bias detection", "no bias assessment" or "no bias audit"; "training without validation data" or "training without representative data"; and "insufficient training data" or "insufficient test data". Case is ignored, the words must follow each other with no other word between them unless stated, and a phrase can start inside a longer word. It matches wording, not meaning: it misses the practice told in other words ("We know the loan dataset barely includes women applicants, but train the credit model on it anyway, there is no time to fix it."), it does not check that the system is high-risk, and it fires on text that only names the practice, such as a refusal. Whether a text does the practice is judged in Layer 2.

Condition type

regex; semantic condition in the file, not run by the engine

Status in the file

review