Preparation for the next AI Act deadline should start from the provision, not from the calendar alone. Identify which systems and operators the dated provision touches, what evidence the provision requires, who owns the work, and what would trigger a reassessment. For the December 2026 milestone that means updating prohibited-practice screening and closing the Article 50(2) transition for the systems it actually covers. For the December 2027 and August 2028 high-risk milestones it means sequencing classification and evidence work so the deadline is a checkpoint, not the start of discovery.
Why does deadline preparation go wrong?
Deadline preparation most often goes wrong when a single date is treated as the whole programme. A team that plans only for 2 December 2027 may overlook duties that are already enforceable, while a team that treats every future date as urgent may spread effort across systems that the dated provision never touches. The AI Act applies in stages and its provisions attach to different operators, systems and uses, so the same date can be a hard application date for one system, a preparation milestone for another and irrelevant for a third. A useful plan therefore names the provision first, then the systems in scope, then the operator roles, and only then the date. That order keeps work proportionate and makes it visible when a change in purpose, model or integration moves a system into a different dated provision.
How do you identify which deadline is next for a system?
For each system, record the intended purpose, the operator role, the market or deployment date, and the provisions already assessed, then test the dated provisions in turn. Start with Article 5 prohibitions and Article 4 literacy, which have applied since 2 February 2025, then Chapter V for general-purpose AI models from 2 August 2025, then Article 50 transparency from 2 August 2026. Next, test the 2 December 2026 changes: the new prohibitions in Article 5(1)(ba) and Article 5(1)(bb), and whether the Article 50(2) transition for a system placed on the market before 2 August 2026 is relevant. Finally, test the high-risk routes: Annex III from 2 December 2027 and Annex I from 2 August 2028, as amended by Regulation (EU) 2026/1744. A system can have more than one next deadline, and the earliest applicable one should drive the immediate work.
For the broader regulatory framework, see EU AI Act timeline.
What should preparation for 2 December 2026 look like?
Preparation for 2 December 2026 is narrow and should stay narrow. Providers should update prohibited-practice screening so the new synthetic-content prohibitions are considered alongside the prohibitions applying since February 2025, examining what a system is intended to do, how it is actually used and what safeguards exist, without reproducing graphic detail or giving examples involving identifiable people. Separately, providers who have relied on the Article 50(2) transition should verify the marking and detection implementation for each system version, confirm the market date, and record testing results and known limitations. Deployers should review their own disclosure duties under other parts of Article 50 rather than assuming a provider's marking work completes their position. The work should finish with a post-deadline review date, because a later model or pipeline change can invalidate an assessment made for the December date.
What should preparation for the high-risk dates look like?
Preparation for 2 December 2027 and 2 August 2028 is longer work and should be sequenced by classification first. The team should map each candidate system to its route, Annex III use case or Annex I product legislation, test the Article 6 conditions including any Article 6(3) question, and record the reasoning before building controls. Provider work then follows the requirement areas in the Act, including risk management, data governance, technical documentation, record-keeping, deployer information, human oversight, accuracy, robustness and cybersecurity, together with the applicable conformity, registration and quality-management processes. Deployer work should be planned in parallel: instructions, oversight assignments, monitoring and log retention. Giving each system an owner, a classification record and a dated evidence plan prevents the deadline from arriving while discovery is still under way.
How do you keep a deadline plan honest as law and guidance change?
A deadline plan stays honest only if it records its sources and its review triggers. Each dated item should carry the provision, the application date, the official source consulted and the date last checked, so an amendment or new guidance can be traced to the items it affects. Regulation (EU) 2026/1744 shows why this matters: it changed the high-risk application dates in Article 113, added new prohibitions and adjusted other provisions, so plans built on the original 2024 dates needed revision. Guidance, standards and codes of practice should be labelled by source and status, and expected material should never be filed as if it were fixed law. When a source changes, the plan should record what changed, which systems were reassessed and who approved the update, rather than silently overwriting the earlier position.
Continue with which AI Act rules to check now in late 2026.
What does this guide not decide?
This guide does not set a deadline for any particular system, does not classify a system as high-risk or prohibited, and does not determine what a specific provider or deployer must do before a given date. Those conclusions depend on the provision, the system's intended purpose and the facts of its deployment, and they can be affected by transitions in Article 111 and by sector legislation that this guide does not analyse. The guide also does not predict future amendments, standards or guidance, and expected material described here should be checked against the current official text before it is relied on. Where a release, contract or conformity decision depends on a deadline, confirm the position with a competent authority or legal adviser and keep the dated reasoning with the decision record.
Frequently asked questions
What is the next AI Act deadline after late October 2026?
The next fixed milestone in the verified timeline is 2 December 2026, covering new prohibitions in Article 5(1)(ba) and Article 5(1)(bb) and the end of a limited Article 50(2) transition. Whether it touches a given organisation depends on the systems and operators involved.
Can one deadline plan cover a whole AI portfolio?
One calendar can list the dates, but the work cannot be identical for every system. Each system needs its own provision, route, operator role and evidence record, because the same date can apply differently across a portfolio.
How often should a deadline plan be reviewed?
At each fixed milestone, when guidance or standards change, and whenever a system's purpose, model, integration or user population changes. A review should record what was checked, the source used and the date, so the next reviewer can see what still stands.
Continue through this AI Act cluster
This article is an informational overview, not a legal determination. Check the current legal text and official implementation guidance for your system and role.
Consolidated Regulation (EU) 2024/1689, as amended ↗AI Act Service Desk implementation timeline ↗