Updated 17 October 2026. This article has been revised to separate the policy timeline from the product argument. See the current AI Act timeline for the category-specific dates and primary reference.
Deferral is a decision, even when it arrives disguised as silence. Moving a piece of AI governance work to next quarter should explain what happens in the meantime.
Make postponement explicit
A deferral record should name the work, the reason for postponement, the exposure that remains and the person accepting the decision. It should also name a review date or a trigger that brings the question back. “Later” is not a trigger.
Consider an illustrative workflow with an untested recovery procedure. The team might restrict the release, use a manual process or delay deployment while it completes the test. Which option is appropriate depends on the consequences and applicable obligations. Pretending the missing test passed is not an option.
Distinguish work that can be deferred from requirements that must be satisfied before a particular use. That assessment needs the relevant expertise. A product deadline does not give a team authority to waive a legal obligation.
Review repeated extensions carefully. If the same task keeps moving, the problem may be ownership, capacity or an unrealistic design rather than scheduling. Another date will not resolve a dependency nobody has understood.
| Decision | Evidence to request |
|---|---|
| Reason | Specific constraint or dependency |
| Interim state | Restriction or compensating measure |
| Return trigger | Owner and review condition |
What to do with this
A clear postponement can be an honest management decision. An invisible postponement quietly transfers the consequences to whoever encounters the failure first.
Record the decision, the responsible owner and the evidence still missing. The decision-record example shows a lightweight way to do that. Where a claim concerns system behaviour, connect it to evaluation evidence. These are proposed product practices, not a finding that a system complies with the law.
Primary timeline reference: European Commission, AI Act. The operational recommendations above are my analysis.
