Technical

data sheet

Record of dataset provenance and characteristics.

How the laws define this

Both regulatory sources agree that a data sheet is a structured document outlining a dataset's origins or provenance, composition, preprocessing steps, and known biases to enable risk assessment. They diverge primarily regarding the specific technical elements required and the target application of the dataset. The NSTDA AI Ethics Guideline defines the record as a metadata document incorporating collection methods, limitations, and suitability for particular tasks to guide dataset reuse. In contrast, the AI and Data Protection Regulation Bill defines it as a provenance report detailing labeling practices and usage restrictions specifically for datasets applied to training or evaluating AI systems. Furthermore, while both require risk assessment, the AI and Data Protection Regulation Bill expands the document's purpose to explicitly include supporting reproducibility and mitigating harms.

Synthesised from the 2 statutory definitions below. Each one is quoted in full, with its source.

How jurisdictions differ

What is distinctive in each definition, binding law in force first.

Jurisdiction	Instrument	Status	What is distinctive
Thailand	NSTDA AI Ethics Guideline	In Force	Adds details on collection and preprocessing methods, known biases or limitations, and suitability for particular tasks to inform dataset reuse and risk assessment.
Ecuador	AI and Data Protection Regulation Bill	Under Review	Adds labeling practices, usage restrictions, and specific intent to support reproducibility, risk assessment, and mitigation of harms for AI training or evaluation datasets.

What it triggers

Under the NSTDA AI Ethics Guideline, it sets out a dataset metadata document to inform dataset reuse and risk assessment.source
In the Ecuador AI and Data Protection Regulation Bill, it would require a dataset provenance report to support reproducibility, risk assessment, and mitigation of harms.source

Definitions (2)

A dataset metadata document detailing provenance, composition, collection and preprocessing methods, known biases or limitations, and suitability for particular tasks to inform dataset reuse and risk assessment.

NSTDA AI Ethics GuidelineDefinition 1 of 2

A dataset provenance report detailing the origins, composition, labeling practices, preprocessing steps, known biases, and usage restrictions of datasets used to train or evaluate AI systems, intended to support reproducibility, risk assessment, and mitigation of harms.

AI and Data Protection Regulation BillDefinition 2 of 2

Not to be confused with

Dataset Provenance

Dataset provenance refers to traceable metadata on origin and composition, while this term covers a fuller document or report including preprocessing, biases, and suitability.

Datasets

Datasets are the curated data collections themselves, while this term is a record or report describing their provenance and characteristics.

Questions and answers

What must a data sheet include under NSTDA guidelines?

It details provenance, composition, collection and preprocessing methods, known biases or limitations, and suitability for particular tasks. This informs dataset reuse and risk assessment.

Does the Ecuador bill mandate data sheets for AI?

The bill would require a dataset provenance report detailing origins, composition, labeling practices, preprocessing steps, known biases, and usage restrictions for datasets used to train or evaluate AI systems.

How do data sheets help with AI risks?

Both definitions emphasize informing risk assessment: NSTDA for reuse and suitability, Ecuador for reproducibility and harm mitigation.