Lifecycle
All stages from design through deployment, operation, and decommissioning.
How the laws define this
Every regulatory standard agrees that an AI lifecycle encompasses the total duration of an AI system's existence, from early design or planning through to eventual decommissioning or retirement, requiring governance, risk assessment, and control measures across each stage. The key divergence lies in the granularity of the defined phases and whether specific technical processes are explicitly named as distinct stages. Concise frameworks, such as Malta — Towards Ethical and Trustworthy AI (Malta's Ethical AI Framework) and the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law (CETS No. 225), group the lifecycle into high-level phases like design, development, use, and decommissioning. Conversely, Strategy: Use of AI systems in the Federal Administration explicitly specifies technical sub-stages, including data collection and processing, model building/adaptation, and testing/validation. Principles and Controls of AI Ethics (SDAIA AI Ethics Principles) similarly expands this scope by detailing conception, testing, and maintenance as dedicated phases requiring continuous compliance assessments.
Synthesised from the 5 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 4 of the 5 instruments that define this term; all are listed under Definitions below.
| Jurisdiction | Instrument | Status | What is distinctive |
|---|---|---|---|
| Council of Europe | Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law (CETS No. 225) | Awaiting Entry | Defines lifecycle plainly as the entire duration of an AI system's existence without specifying distinct stages or additional requirements. |
| Saudi Arabia | Principles and Controls of AI Ethics (SDAIA AI Ethics Principles) | In Force | Adds conception and retirement to the phase sequence and describes applying principles, controls, and assessments at each stage for compliance. |
| Switzerland | Strategy: Use of AI systems in the Federal Administration | In Force | Specifies a defined set of seven stages for governance and sets out documentation, risk assessment, and control measures at each. |
| Malta | Malta — Towards Ethical and Trustworthy AI (Malta's Ethical AI Framework) | Adopted | Orients the phase sequence explicitly toward governance, risk management, and control activities across the system's operational life. |
What it triggers
- Under the Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law, iterative risk and impact assessment processes are required across all lifecycle stages.source
- Saudi Arabia's Principles and Controls of AI Ethics describes the lifecycle as phases from conception through retirement and applying principles, controls, and assessments at each stage for compliance and risk mitigation.source
- Switzerland's Strategy: Use of AI systems in the Federal Administration sets out specific documentation, risk assessment and control measures under each lifecycle stage.source
Definitions (5)
The entire duration of an AI system's existence, including design, development, use, and decommissioning.
The sequence of phases through which an AI system passes—design, development, deployment, monitoring, and decommissioning—used to orient governance, risk management, and control activities across the system's operational life.
The sequence of phases covering conception, design, development, testing, deployment, monitoring, maintenance and retirement of an AI system. The lifecycle framing mandates that principles, controls, and assessments be applied at each stage to ensure continuous compliance and risk mitigation.
All stages of an AI system including design, research and development, testing, deployment, operation, maintenance and monitoring, through to decommissioning; the Convention requires iterative risk and impact assessment processes across these stages.
The defined set of stages applied to AI systems for governance and oversight: plan & design; data collection and processing; model building/adaptation; testing/validation; deployment; operation/monitoring; and decommissioning; each stage requires specific documentation, risk assessment and control measures under the strategy.
Not to be confused with
AI lifecycle refers specifically to stages of AI from planning through decommissioning, while lifecycle is a broader term encompassing various phase sequences across instruments.
Activities within the lifecycle names the stages themselves from design to decommissioning, whereas lifecycle denotes the overall sequence or set of those stages.
Questions and answers
What stages are included in the AI lifecycle under the Council of Europe AI Convention?
It covers all stages including design, research and development, testing, deployment, operation, maintenance and monitoring, through to decommissioning. The Convention requires iterative risk and impact assessment processes across these stages.
What does Saudi Arabia's AI Ethics Principles say about the lifecycle?
It defines the lifecycle as phases from conception through retirement and describes principles, controls, and assessments at each stage for compliance and risk mitigation.
How does Switzerland structure the AI lifecycle in its federal strategy?
It sets out seven stages: plan & design; data collection and processing; model building/adaptation; testing/validation; deployment; operation/monitoring; and decommissioning. Each describes documentation, risk assessment, and controls.
Related Terms
activities within the lifecycle
Stages of an AI system from design to decommissioning....
AI lifecycle
Stages of AI from planning through decommissioning....
AI life cycle
Stages from scoping through decommissioning....
Life cycle
The sequence of phases governing an AI system from design to decommissioning....
cycle of life
Stages from R&D through deployment and maintenance....