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.
The move from August 2026 to December 2027 is sixteen months, not eighteen as an earlier version of this article stated. The more interesting question is what happens to the budget when the pressure eases.
Protect the work from calendar drift
A deadline extension can create room for better implementation. It can also make essential work lose its funding to the next visible feature. Neither outcome follows automatically. The organisation chooses through its allocation of people, time and authority.
For an illustrative quarterly review, separate the cost of operating a control from the cost of preparing for a deadline. Logging, access management and incident handling can remain useful while a legal transition continues. Whether they are required in a particular case needs its own assessment.
Ask what will stop working if a budget line is removed. If the answer is “we will no longer rehearse recovery” or “nobody will maintain the system inventory”, the trade-off should be recorded plainly. Calling the cut an efficiency does not change its effect.
| Decision | Evidence to request |
|---|---|
| Retain a control | Named purpose and operational benefit |
| Defer work | Consequence and interim measure |
| Reduce budget | Explicit owner acceptance of the trade-off |
When you move a deadline, you don’t just extend the timeline. You release the pressure that was making the work happen.
It’s the middle tier that gets the most relief: companies that had started the work, were 40 or 50 percent done, and now have political cover to deprioritize. That work doesn’t continue at the same pace. It competes with shipping new features, cutting costs, and hitting quarterly targets. More time isn’t always more readiness. Often it’s just more delay.
What to do with this
A sensible plan can use additional time without pretending the work became free. Keep milestones tied to evidence and review the actual result. A revised date should trigger a planning decision, not automatically delete the plan.
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.
