System retirement is the planned decommissioning of an AI system from production when it is no longer needed, no longer performs adequately, or cannot be operated safely. It is a lifecycle activity, not simply switching off a model: teams should identify system owners, affected workflows, dependent services, and any records or data that must be retained.
In cyber security and AI governance, a controlled retirement helps prevent abandoned systems, forgotten credentials, exposed data, and unsupported components from remaining accessible. Practical steps can include disabling integrations and accounts, revoking keys, removing infrastructure, preserving required audit records, and notifying users of workflow changes. Organizations should also verify that downstream processes have a safe replacement or fallback and confirm that the retired system is no longer processing data.