title: “Retirement Schema” description: “Machine-readable representation of AI system retirement activities.”
Retirement Schema
The Retirement Schema provides the machine-readable structure for representing the controlled retirement of an AI system and the governance activities required to transition it from active operation to a retired state. Retirement is a lifecycle governance activity. It should preserve appropriate accountability, traceability, evidence, residual-risk information, and required records after active use of the AI system has ended.Purpose
The Retirement Schema supports structured recording of: * retirement decisions * retirement rationale * system retirement status * decommissioning activities * access removal * dependency handling * data and record disposition * residual risks * contractual or governance obligations * stakeholder communications * evidence retention * post-retirement actions * responsible owners * verification * closure The schema provides the structural representation of a retirement record. It does not prescribe a single technical decommissioning method.Retirement triggers
An AI system may enter retirement for different reasons, including: * end of intended use * replacement by another system * unacceptable risk * business or operational changes * changes in applicable requirements * inability to maintain required controls * material changes to the operating environment * technology obsolescence * organizational decisions * contractual or strategic changes The reason for retirement should be recorded so that the resulting governance decision remains understandable and traceable.Retirement decision
Retirement should be governed as a formal lifecycle decision where appropriate. The decision may consider: * current system status * classification * active risks * residual risks * control status * outstanding incidents * outstanding changes * assurance findings * contractual obligations * regulatory or legal requirements * evidence-retention requirements * affected stakeholders * system dependencies * replacement or transition arrangements Where approval is required, the retirement decision should be linked to the applicable approval record.Retirement activities
Decommissioning may involve activities such as: * disabling system access * removing or disabling deployment infrastructure * terminating integrations * handling dependent systems * revoking credentials * transferring ownership * preserving required records * disposing of data where permitted and appropriate * communicating retirement to stakeholders * documenting the final operational state The applicable activities depend on the nature and architecture of the AI system.Residual risk
Retirement does not necessarily eliminate all governance risk. Residual risks may remain after active operation ends, including: * retained data * retained models or artifacts * contractual obligations * regulatory obligations * archived evidence * downstream dependencies * historical decisions * ongoing stakeholder impacts * security or privacy considerations Relevant residual risks should remain traceable to the retirement record and managed according to applicable governance requirements.Evidence and records
Retirement should preserve evidence necessary to demonstrate that required retirement activities were completed. Evidence may include: * retirement approval * decommissioning records * access-removal records * data disposition records * dependency records * communications * final monitoring results * outstanding risk decisions * assurance results * verification records Records should be retained according to applicable organizational, contractual, legal, regulatory, security, privacy, and records-management requirements.Post-retirement state
The retirement record should establish the resulting lifecycle state of the AI system. Where appropriate, the organization should distinguish between: * active operation * retirement in progress * retired * decommissioned * archived * replaced The precise status vocabulary should follow the applicable AIGO schema and organizational governance definitions.Traceability
Retirement should remain connected to the AI system and relevant lifecycle records. Relationships may include: * AI system * classification * risks * controls * assessments * approvals * monitoring * incidents * changes * assurance * evidence * improvements A representative lifecycle relationship is: AI system → retirement decision → decommissioning → verification → post-retirement records This preserves governance history even after active operation has ended.Relationship to evaluation
Retirement information may participate in machine-readable governance evaluation. The AIGO evaluation architecture separates governance records from evaluation logic: governance records → rules → evaluation → result Evaluation may be used to check defined retirement conditions, such as required approvals, evidence, status transitions, or completion of specified governance activities. Automated evaluation should support rather than replace governance judgment concerning whether retirement requirements have been appropriately satisfied.Relationship to continuous improvement
Retirement can provide useful inputs to continual improvement. Organizations may review retirement outcomes to identify: * recurring lifecycle weaknesses * inadequate documentation * dependency-management problems * evidence-retention issues * control weaknesses * governance process improvements * lessons for future AI system deployments These observations can be connected to Improvement records where appropriate.Machine-readable governance
The Retirement Schema forms part of the AIGO machine-readable governance layer. It can establish relationships with: * AI systems * risks * controls * assessments * approvals * monitoring * incidents * changes * assurance * evidence * management reviews * improvements This allows retirement to remain part of the complete AI governance lifecycle rather than being treated as a separate technical decommissioning activity.Validation
Retirement records participate in the broader AIGO validation and traceability model. Repository validation may include checks relating to: * schema structure * references * traceability * evidence relationships * framework consistency * document integrity * repository health * machine-readable governance artifacts A successful validation result provides evidence about the integrity of the validated repository or artifact scope. It does not by itself establish legal compliance, certification, accreditation, or regulatory conformity.Canonical source
Machine-readable schema:schemas/13-retirement/13-AIGO-Retirement-Schema-v0.1.json
Schema documentation:
schemas/13-retirement/13-AIGO-Retirement-Schema-Documentation-v0.1.md