EU AI Act deadlines and compliance calendar

Article 113 is the anchor for the AI Act calendar. The Regulation applies from 2 August 2026 as the general application date, but Article 113 brings selected chapters into application earlier and delays Article 6(1) product-safety high-risk classification until 2 August 2027.

Use one calendar row per application date. Each row should identify the legal provision, affected systems or models, accountable owner, required evidence, and the internal readiness date that comes before the legal application date.

The first operating deadline is not a paperwork date. It is the point when Chapter I and Chapter II apply, so the calendar should force two workstreams: an AI literacy owner for staff and other persons operating or using AI systems on the organisation's behalf, and a prohibited-practices owner for product and deployment screening.

The useful evidence is concrete: training audience, AI system context, role-specific literacy material, product-screening notes, and a release gate showing that prohibited-practice review happened before placing on the market, putting into service, or use.

The 2 August 2025 row should separate GPAI model-provider obligations from ordinary AI-system deployment work. Article 113 brings Chapter V into application on this date, while Commission GPAI guidance states that GPAI provider obligations enter into application on 2 August 2025.

For organisations building on third-party GPAI models, the calendar should still assign a procurement and product owner. They need model documentation requests, downstream integration records, supplier contact points, and evidence that the product team knows whether the organisation is only deploying a model, integrating it into an AI system, or significantly modifying it.

The 2 August 2026 row should be the broad readiness gate for AI Act obligations that are not already in force under Article 113. Product, legal, security, privacy, and operations owners should use it to close inventory gaps and map each system to its role, risk category, deployment context, and evidence owner.

The 2 August 2027 row is narrower but important for product teams. Article 113 applies Article 6(1) and corresponding obligations from that date; Article 6(1) covers AI systems used as safety components of products, or AI systems that are themselves products, where the covered product route requires third-party conformity assessment under Annex I Union harmonisation legislation.

Article 111 is not a general grace period for every AI system. It creates specific transitional treatment for large-scale IT systems, pre-2 August 2026 high-risk systems, public-authority high-risk systems, and GPAI models placed on the market before 2 August 2025.

Keep these rows separate from Article 113 because the owner question is different: not just when a chapter applies, but whether an already-placed system or model is grandfathered, changed, public-authority-facing, or subject to a later compliance deadline.

A useful AI Act calendar is more than a date list. Each row should be tied to a system or model inventory view so product and compliance teams can see which obligations are approaching, which evidence is missing, and who can unblock the work.

Do not validate unsupported future guidance dates in the calendar. If a date is not in the cited source material, leave it out and track the dependency as a watch item rather than a legal milestone.