By late 2026, several major parts of the EU AI Act are already applicable. These include the general provisions and definitions, AI literacy duties, most prohibited AI practices, general-purpose AI model obligations, governance provisions and Article 50 transparency requirements. Enforcement powers also began applying on 2 August 2026 for applicable rules. Some important provisions still have later dates, including new prohibitions and a transition under Article 50(2) on 2 December 2026, Annex III high-risk rules on 2 December 2027 and Annex I product-related high-risk rules on 2 August 2028, as amended by Regulation (EU) 2026/1744.
Which AI Act Rules Were Already Applicable in 2026?
By late 2026, the AI Act is not a single future obligation but a set of provisions that have reached their application dates at different times. The general provisions and definitions, AI literacy duties, the original prohibited practices, general-purpose AI model obligations, governance provisions and Article 50 transparency requirements are already applicable, and enforcement powers began applying on 2 August 2026 for the provisions concerned. That position must be read provision by provision: an organisation's duties depend on its role, its systems and its uses, and a duty that is already enforceable for one system may still be in preparation for another. The practical task is therefore to separate what is already applicable from what has a later date, and to record the provision, the date and the evidence for each system, rather than treating 2026 as either a year of full application or a year in which nothing is yet due.
What Began Applying on 2 February 2025?
On 2 February 2025, the first substantive parts of the Act began applying: the general provisions and key definitions, the AI literacy requirement in Article 4 for providers and deployers, and the original prohibited practices in Article 5. The application of the definitions mattered because the rest of the Act is built on them: whether something is an AI system, who is a provider or a deployer, and what an intended purpose is, all depend on the general provisions. AI literacy required providers and deployers to take measures to ensure a sufficient level of literacy for staff and others operating or using AI on their behalf, taking context into account. The prohibited practices became an immediate screening issue for every use case, since a prohibited use must not be placed on the market, put into service or used at all. That first milestone was therefore operational from the start, not merely preparatory.
For the broader regulatory framework, see EU AI Act timeline.
What Began Applying on 2 August 2025?
On 2 August 2025, the Act entered its general-purpose AI and governance stage. Obligations for providers of general-purpose AI models under Chapter V became applicable, including technical documentation, information for downstream providers, a copyright policy and a public summary of training content, with additional duties for models presenting systemic risk. At the same time, the governance provisions had to be in place: the AI Office, the European Artificial Intelligence Board, the Scientific Panel and the Advisory Forum at Union level, and national competent authorities designated by the Member States, together with national rules on penalties. That combination mattered for two audiences at once: model providers faced substantive duties, and every operator gained a supervisory structure to engage with. The date did not bring forward transparency or high-risk duties, which kept their later dates, but it did make the model layer of the Act a live compliance question.
What Became Applicable on 2 August 2026?
On 2 August 2026, the majority of the remaining rules began applying, including the Article 50 transparency requirements, and measures supporting innovation became operational. Enforcement also started for applicable rules at national and Union level, covering prohibitions, general-purpose AI models, transparency and AI literacy within the respective competences of the AI Office and national authorities. Article 50 duties are targeted rather than universal, so the August 2026 date required providers and deployers to work through each system and output type, distinguishing provider marking duties from deployer disclosure duties and checking the conditions and exceptions in the provision. A limited transition for certain providers of systems already placed on the market before 2 August 2026 continued until 2 December 2026 for the Article 50(2) marking and detection obligation, and that transition should not be read as postponing Article 50 generally or any other applicable duty.
Which Important AI Act Rules Were Not Yet Applicable?
Several important rules still had later dates in late 2026. On 2 December 2026, new prohibitions in Article 5(1)(ba) and Article 5(1)(bb), concerning AI systems used to generate or manipulate non-consensual intimate material and material within Directive 2011/93/EU, were scheduled to apply, together with the end of the Article 50(2) transition for certain pre-existing systems. Chapter III Sections 1 to 3, the high-risk requirements, were scheduled to apply from 2 December 2027 for Annex III systems under Article 6(2) and from 2 August 2028 for Annex I product systems under Article 6(1), following the amendments made by Regulation (EU) 2026/1744. Those later dates do not postpone the duties already applicable: a system can have a future high-risk date while already facing literacy, prohibition, general-purpose AI or transparency duties today, and the planning record must carry both positions at once.
Continue with what changed on 2 August 2025.
How Do You Determine What Applies to Your Organisation Now?
Determining what applies now starts with the system and the model, not with a checklist of dates. Identify each AI system and general-purpose AI model, record its intended purpose and the people it affects, and identify the operator role for each organisation involved: provider, deployer, importer, distributor or authorised representative. Screen each use for prohibited practices under Article 5, check whether a general-purpose AI model relationship creates Chapter V duties or downstream information needs, and test the Article 50 transparency duties for the interaction and output types concerned. Then identify any later high-risk milestone, Annex III or Annex I, that the system may meet, and record the classification reasoning separately from the duties already applicable. Where a question is borderline, record the uncertainty, the source consulted and the reviewer, and seek qualified review rather than treating an unresolved classification as a conclusion.
What Should Organisations Document in 2026?
In 2026, an organisation should be able to show, for each system, the inventory entry, the intended purpose, the operator role and its rationale, the application date for each provision assessed, the prohibition screening, the AI literacy measures taken, the general-purpose AI dependencies and the transparency decisions made. Each entry should carry the source provision, the decision rationale, the reviewer and the review date, so a change in the system, its purpose or the law can trigger reassessment. Documentation should be kept in a form that can be produced to an authority, shared with a deployer or provider in the chain, and understood by a new reviewer who was not involved in the original decision. A record that states only a conclusion, without the provision, the evidence and the date, does not support a compliance claim, and a record that is not updated when the system changes will gradually describe a system that no longer exists.
Frequently asked questions
Is the EU AI Act already enforceable in 2026?
Yes. Enforcement powers began applying on 2 August 2026 for provisions that are already applicable, including rules concerning prohibited practices, transparency, AI literacy and general-purpose AI models. Enforcement follows each provision's application date, so later high-risk duties are not yet enforceable on the same basis.
Are the high-risk AI rules already fully applicable in 2026?
No. Under the current amended timeline, the Annex III high-risk rules apply from 2 December 2027 and the Annex I product-related high-risk rules apply from 2 August 2028, as amended by Regulation (EU) 2026/1744. Other provisions are already applicable and should not be treated as postponed by those later dates.
Do companies need to wait for the high-risk deadlines before doing AI Act work?
No. Several obligations are already applicable, and organisations can also use the remaining transition period to classify systems and prepare for later high-risk requirements. Classification, documentation, testing and oversight design take lead time, and the later dates are checkpoints for that work, not starting points.
What does this article not decide?
This article does not decide whether any particular system is prohibited, high-risk or subject to a specific duty, and it does not tell a reader what they must do for a specific system. Those questions turn on the system's intended purpose, the exact legal wording and the facts of deployment, and should be confirmed with a competent authority or legal adviser where a decision depends on a contested interpretation.
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 ↗European Commission AI Act overview ↗