How the EU AI Act's Phased Timeline Works
The EU AI Act (Regulation (EU) 2024/1689) entered into force on 1 August 2024 and applies in stages rather than all at once. Prohibited AI practices and AI literacy obligations became applicable on 2 February 2025. General-purpose AI (GPAI) model obligations, along with the governance structures that oversee them, became applicable on 2 August 2025. The remaining and most operationally demanding provisions apply during 2026 and 2027. Governance teams that have not yet mapped their AI system inventory against the Act's risk categories should treat this as the immediate prerequisite to any deadline planning, since the applicable date for a given system depends entirely on how it is classified.
Key Dates Through 2027
- High-Risk Annex III Systems and Broad Enforcement: Most remaining provisions of the Act become applicable, including obligations for high-risk AI systems listed in Annex III, transparency duties, and the bulk of penalty and market surveillance provisions. This is 24 months after entry into force.
- High-Risk Annex I Safety-Component Systems: Obligations apply to high-risk AI systems that function as safety components of products already covered by EU harmonization legislation, such as machinery or medical devices. This is 36 months after entry into force.
What the 2 August 2026 Deadline Requires
Providers of high-risk systems under Annex III must have a documented risk management system that is maintained across the system lifecycle, as required under Article 9. Article 11 and Article 12 require technical documentation and automatic logging capable of supporting traceability of system operation, meaning the system must be able to reconstruct what happened, when, and under what inputs. Article 14 requires human oversight measures that allow a human operator to understand, monitor, and intervene in system outputs, not merely review outcomes after the fact. Conformity assessment, whether through self-assessment or a notified body depending on system category, must be completed before the system is placed on the market. Because obligations differ by role, providers, deployers, importers, and distributors each carry distinct accountability under the Act, and governance programs should assign ownership accordingly rather than treating compliance as a single centralized function. Auditability requirements here are directly technical: logging and record-keeping must capture inputs, outputs, and decision context in a form that can withstand review by a national market surveillance authority, and human oversight interfaces need real-time intervention capability rather than static dashboards.
What the 2 August 2027 Deadline Requires
The 2027 deadline applies specifically to high-risk AI systems that are safety components of products regulated under existing EU harmonization legislation, such as machinery, medical devices, or similar product categories under Annex I. These systems must meet the same underlying risk management, documentation, logging, and human oversight requirements as Annex III systems, but conformity assessment must be aligned with the existing sectoral CE-marking regime for that product category rather than assessed as a standalone AI obligation. Organizations building or embedding AI into regulated physical products should treat this as a distinct workstream from Annex III readiness, since the assessment pathway, timing, and responsible bodies differ. Waiting until 2027 to begin this alignment work is a significant risk given that product-safety conformity processes typically require lead time measured in months, not weeks.
Common Questions on the 2026-2027 Timeline
Who enforces the AI Act during this period?
National market surveillance authorities, designated by each Member State, hold primary enforcement responsibility for most high-risk systems. The EU AI Office holds oversight responsibility specifically for GPAI model obligations, creating a dual regulatory structure that governance teams need to track separately.
What are the penalty tiers under the Act?
Violations of prohibited practices can result in fines up to €35 million or 7% of worldwide annual turnover. Other high-risk obligation violations can result in fines up to €15 million or 3% of turnover. Supplying incorrect or misleading information to authorities can result in fines up to €7.5 million or 1% of turnover.
Could these dates change before 2026?
EU digital simplification proposals, sometimes referred to as a Digital Omnibus initiative, have raised the possibility of adjustments to some AI Act deadlines. As of this writing, the legislative status of any such change has not been finalized, and organizations should verify current dates independently before finalizing readiness timelines.
Do GPAI obligations from 2025 affect 2026 high-risk readiness?
Yes. Downstream deployers integrating GPAI models into high-risk systems need the provider's technical documentation and, for models with systemic risk, evaluation and incident data, to complete their own risk management records ahead of the 2026 deadline.