The EU AI Act: milestones to verify

The EU AI Act does not begin on a single enforcement date. It entered into force on 1 August 2024 and applies in stages. According to the European Commission's implementation timeline, prohibited AI practices, the definition provisions, and AI literacy duties applied from 2 February 2025. Governance rules and obligations for general-purpose AI models applied from 2 August 2025.

What changes on 2 August 2026

Most of the Act becomes generally applicable on 2 August 2026. This includes the transparency duties in Article 50 for certain AI systems, such as specified disclosure duties for people interacting with AI and labelling requirements for certain synthetic content. Applicable duties depend on the actor, system, and use case.

This date does not newly activate the prohibited-practice rules; those already applied from February 2025. Nor does it mean that every high-risk system requirement starts on this date. The Commission's current implementation timeline places later milestones on specified categories of high-risk systems, including transitional dates affected by the Digital Omnibus process.

High-risk systems and transitional dates

The legal classification of a high-risk system depends on the system's intended purpose and whether it falls within the categories set out in the Act. The Commission's current timeline identifies later application dates for specified Annex III and Annex I high-risk systems. Organisations should check the official timeline and the final status of transitional measures rather than relying on a single summary date.

Operational interpretation

The following is Verinox AI's operational interpretation, not a statement of law: organisations should maintain an AI system inventory, record intended purposes and responsible actors, screen for prohibited practices, identify applicable transparency duties, and document why a system is or is not treated as high risk. Vendor contracts should support access to the information needed for the organisation's own duties.

Practical verification steps

1. Inventory AI systems and roles. Record whether the organisation is acting as provider, deployer, importer, distributor, or another actor.
2. Check prohibited practices. Confirm that controls introduced for the February 2025 milestone remain effective.
3. Assess Article 50 transparency. Identify interfaces and synthetic content that require disclosure or machine-readable marking from August 2026.
4. Classify potential high-risk systems. Map each intended purpose to the official categories and record the reasoning and applicable transition date.
5. Recheck primary sources. Implementation dates can be affected by adopted amendments and transitional measures.

This article provides general information, not legal advice. The Regulation, adopted amendments, and European Commission guidance remain authoritative.