EU AI Act deadlines: 2026 to 2028

What already applies

The regulation entered into force on 1 August 2024. Prohibited-practice and AI-literacy provisions applied from 2 February 2025, while governance and general-purpose AI provisions began applying from 2 August 2025. Organisations should therefore already have addressed these areas where relevant.

• Record prohibited-practice screening
• Evidence proportionate AI-literacy measures
• Identify any GPAI-model-provider role
• Keep the source and review date with each conclusion

2 August 2026

Most remaining provisions apply from this date, including Article 50 transparency obligations. Some high-risk timing was amended, so a system-specific plan must distinguish transparency, operator duties, enforcement and the later high-risk dates.

• Map direct AI interaction disclosures
• Review synthetic-content and deepfake controls
• Confirm emotion-recognition and biometric notices
• Retain implementation and accessibility evidence

2 December 2027 and 2 August 2028

The principal high-risk dates now distinguish listed stand-alone use cases and systems that are safety components of regulated products. Confirm the current official text and transition rules before fixing a delivery date.

• Prioritise likely Annex III use cases
• Identify regulated-product dependencies
• Back-plan technical documentation and conformity work
• Set reassessment triggers for purpose and model change

Turn dates into an operating plan

For every system, retain the applicable date, source, rationale, accountable owner, evidence plan and internal target. Internal dates should allow time for supplier information, technical changes, training and independent review.