Data
Public and private data assets, emphasizing openness and interoperability.
How the laws define this
Regulators broadly align on viewing data as information, records, files, or raw facts that are processed, stored, transmitted, or used to train, test, and operate artificial intelligence systems. However, these regulatory texts diverge significantly regarding whether data is defined primarily by personal linkage or by general format and technical scope. The Circular Interpretativa SERNAC — Protección de los consumidores frente al uso de sistemas de inteligencia artificial en las relaciones de consumo focuses specifically on formats linked to identifiable persons, such as behavioral, preference, and inferred consumer details. In contrast, the Advisory Council on the Ethical Use of AI and Data explicitly includes both personal and non-personal datasets across the lifecycle. Other instruments define scope by physical or digital format rather than personal identifiability; the Data Security Law of the People's Republic of China broadly covers any record of information in electronic or other forms, whereas Disegno di legge n.1146 - Disposizioni e deleghe al Governo in materia di intelligenza artificiale explicitly incorporates audio and visual recordings.
Synthesised from the 8 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. Showing 7 of the 8 instruments that define this term; all are listed under Definitions below.
| Jurisdiction | Instrument | Status | What is distinctive |
|---|---|---|---|
| China | Data Security Law of the People's Republic of China | In Force | Expansively covers any record of information in electronic or other forms as baseline for scope, obligations, and classification. |
| Italy | Parliamentary bill n.1146 - Provisions and delegations to the Government regarding artificial intelligence | In Force | Expands EU AI Act terminology to expressly include audio and visual recordings as inputs and assets across the AI lifecycle. |
| Iran | Computer Crimes Law | In Force | Limits to information, files, records or content processed, stored, transmitted or received via computer or telecommunications systems, distinguishing subcategories by sensitivity. |
| Singapore | Advisory Council on the Ethical Use of AI and Data | In Force | Highlights both personal and non-personal datasets across the AI lifecycle with concerns for quality, provenance, and PDPA applicability. |
| Chile | Interpretative Circular on consumer protection against the use of artificial intelligence systems in consumer relationships | In Force | Broadly reads to include numeric, alphabetic, photographic, acoustic or other formats linked to identifiable persons plus behavioral, preference and inferred consumer information. |
| Argentina | National Artificial Intelligence Plan | Adopted | Emphasizes public and private data assets like open public datasets, standardised metadata, and controlled access models matching the page's one-line definition. |
| British Columbia, Canada | Draft Artificial Intelligence Responsible Use Principles | Draft | Limits to raw facts, figures, and statistics that AI systems process to identify patterns, make predictions, or generate outputs, stressing quality and representativeness. |
What it triggers
- In China, data falling under this definition triggers risk management and transfer controls for data processors.source
- Under Parliamentary bill n.1146 in Italy, data used for training, validation, inputs and outputs of AI models triggers national rules and criminal penalties where the text applies.source
- Under Iran's Computer Crimes Law, data serves as the basic object of protection framing offences like unauthorized access and disclosure with applicable penalties.source
Definitions (8)
Encompasses public and private data assets, including open public datasets, standardised metadata, and governance models for controlled access to private sector data. The Plan highlights interoperable, open data standards and alignment with existing data protection law for responsible use.
The statute adopts the EU AI Act terminology but expressly expands the notion of 'data' at the national level to cover a wide range of inputs and assets used across the AI lifecycle, explicitly including audio and visual recordings; this definition governs data used for training, validation, inputs and outputs of AI models where the national text applies.
Includes both personal and non‑personal datasets used throughout the AI lifecycle (training, validation, testing, and operation), and highlights concerns around data quality, provenance and applicability of data protection obligations such as the PDPA when personal data are involved.
Any record of information in electronic or other forms, covering a broad range of factual or digital records regardless of format. The DSL uses this expansive definition as the baseline for scope, obligations and classification under the law.
Any information, files, records or content that are processed, stored, transmitted or received via computer or telecommunications systems; the statute uses this term as the basic object of protection and further distinguishes subcategories by sensitivity. The law’s definition frames offences (e.g., unauthorized access, disclosure) and applicable penalties.
Data is read broadly to include numeric, alphabetic, photographic, acoustic or other formats linked to identifiable persons (consistent with Law N°19.628), and explicitly covers behavioral, preference and inferred consumer information collected or produced by AI systems.
Defined broadly to include datasets generated by the public sector, industrial sensors, healthcare records, and other domain sources that enable AI training and inference; the Strategy frames data availability, sharing frameworks, and privacy-protecting data pipes as essential infrastructure for AI social implementation. It highlights both the value of such data for algorithm development and the need for governance, privacy safeguards, and platform specifications.
Data refers to the raw facts, figures, and statistics that AI systems process and learn from to identify patterns, make predictions, or generate outputs. The quality and representativeness of data are crucial for the fairness and reliability of AI systems.
Not to be confused with
Datasets are curated, annotated collections of data for model training/testing, whereas data is the broader term encompassing raw or uncurated information across instruments.
Data governance and data quality refer to criteria and practices for dataset provenance, labelling and curation, while data is the underlying information subject to those practices.
Questions and answers
What data triggers obligations under China's Data Security Law?
Any record of information in electronic or other forms triggers risk management and transfer controls for data processors in China. The DSL uses this expansive definition as the baseline for scope, obligations and classification.
Does Italy's AI bill expand the definition of data beyond the EU AI Act?
Yes, it expressly expands to cover a wide range of inputs and assets used across the AI lifecycle, explicitly including audio and visual recordings. This governs data for training, validation, inputs and outputs of AI models.
What does data mean for AI ethics in Singapore?
It includes both personal and non-personal datasets used throughout the AI lifecycle, highlighting concerns around data quality, provenance and PDPA applicability when personal data are involved. The Advisory Council guides on responsible deployment.
Related Terms
data policy
Rules and measures for dataset sharing and governance....
dataset
Structured collections used for training and validation of models....
Datasets
Curated, annotated collections of data for model training/testing....
data as an enabler
Datasets and infrastructure that permit AI development and deployment....
Data governance and data quality
Criteria and practices for dataset provenance, labelling and curation....