Overview
ISO/IEC 8183:2023 - Information technology - Artificial intelligence - Data life cycle framework defines a structured data life cycle for AI systems. The standard identifies the stages and associated actions for data processing across an AI system’s lifetime - from idea conception through acquisition, preparation, model building, deployment, operation and decommissioning. It is applicable to all organizations that use data in AI development and use, and explicitly does not prescribe specific services, platforms or tools.
Keywords: ISO/IEC 8183:2023, data life cycle, AI, artificial intelligence, data governance
Key topics and requirements
The document organizes data activities into 10 stages, each with recommended tasks and considerations:
- Stage 1 - Idea conception: define the problem and align it to business objectives.
- Stage 2 - Business requirements: scope, feasibility, compliance, ethics and governance requirements.
- Stage 3 - Data planning: data scope, sources, volume, formats, licensing, privacy, storage and retention.
- Stage 4 - Data acquisition: sourcing, consent, contracts, formats (JSON, XML, streaming, IoT).
- Stage 5 - Data preparation: decryption, cleaning, bias mitigation, feature engineering, normalization, labeling, enrichment.
- Stage 6 - Building a model: model development, verification and validation (internal).
- Stage 7 - System deployment: move from model to system-level deployment and integration.
- Stage 8 - System operation: ongoing monitoring, maintenance and validation of the running system.
- Stage 9 - Data decommissioning: secure deletion, archiving or repurposing of data.
- Stage 10 - System decommissioning: retire the system irrespective of data disposition.
Other notable elements:
- Emphasis on data quality, security, privacy (including DPIA considerations), and ethical outcomes (fairness).
- Encourages mapping life cycle processes to stages and using feedback paths between stages for iterative improvement.
- References ISO/IEC 22989 for AI terminology and other related standards for dataset and life-cycle details.
Keywords: data planning, data preparation, data acquisition, data decommissioning, data governance
Applications
ISO/IEC 8183:2023 is practical for:
- Developing organizational policies for AI data governance and lifecycle management.
- Guiding data scientists and ML engineers on preprocessing, labeling and feature engineering best practices.
- Informing IT, security and privacy teams on data handling, storage and secure deletion requirements.
- Supporting compliance, risk and ethics assessments across AI projects.
- Framing procurement and contractual requirements when acquiring third‑party datasets.
Keywords: AI lifecycle, data governance, ML engineers, data quality
Who should use this standard
- AI program leads, data scientists and ML engineers
- IT and data governance teams
- Legal, compliance and privacy officers
- Product managers and system architects at organizations of any size
Related standards
Keywords: ISO AI standards, AI data lifecycle, ISO/IEC 8183:2023