The next deadline is 2 August 2026
That is the general application date, and it is the big one. If you provide or deploy an AI system that lands in Annex III, the conformity assessment, registration, logging, and human oversight requirements are enforceable from that day, and national authorities can start acting on them straight away.
Complete AI Act Timeline
The staging is deliberate. Practices the EU considers unacceptable were banned first, model-level rules came next, and the heavier engineering obligations on high-risk systems were given a two-year runway.
13 June 2024
European Parliament and Council adopt Regulation (EU) 2024/1689, the first horizontal AI law anywhere.
12 July 2024
The Regulation is published in the Official Journal of the European Union.
1 August 2024
Entry into force, twenty days after publication. The clock starts on every deadline below.
2 February 2025
The Article 5 prohibitions bite. Social scoring, untargeted facial scraping, emotion recognition at work and in schools, and the rest of the banned list have to be switched off, not merely documented.
2 May 2025
The European AI Office publishes the Codes of Practice for general-purpose AI models.
2 August 2025
Chapter V obligations for general-purpose AI providers start. The AI Office, the AI Board, and the Scientific Panel become operational, and Member States put their penalty regimes in place.
2 August 2026
General application. Annex III high-risk requirements become enforceable: conformity assessment, EU database registration, CE marking, human oversight, logging, and fundamental rights impact assessments where they apply. Every Member State must have a working AI regulatory sandbox.
2 August 2027
Annex I kicks in for AI embedded as a safety component in products already covered by EU harmonised safety law, including medical devices, vehicles, aviation, and machinery. General-purpose models that were on the market before August 2025 must be fully compliant by this date.
2 August 2030
High-risk systems operated by public authorities that were already in service before August 2026 have to be brought into compliance.
2 August 2028 and onward
The Commission runs its review cycle, revisiting the Annex III list, the Article 50 transparency thresholds, and enforcement practice roughly every four years.
What to do now, in July 2026
General application is days away and the Annex I deadline is a year behind it. Sequence the work like this.
Before 2 August 2026
- • Finish the inventory. You cannot classify systems you have not found, and shadow deployments are where the surprises live.
- • Confirm which systems fall in Annex III, and document any Article 6(3) derogation before you rely on it.
- • Register high-risk systems in the EU database and complete conformity assessment.
- • Turn on automatic logging and name the humans who hold oversight authority.
- • Run fundamental rights impact assessments where Article 27 applies, which covers public bodies plus banking and insurance.
Through to August 2027
- • Fold AI conformity work into existing product certification if you ship AI inside regulated hardware.
- • Bring general-purpose models that predate August 2025 up to full Chapter V compliance.
- • Keep post-market monitoring running and report serious incidents rather than filing them.
Standing obligations
- • Keep the inventory current as teams adopt new models and agents.
- • Re-run classification when a system changes materially, since a redesign can move it into scope.
- • Retain logs for the statutory period and keep them queryable.
Transitional provisions worth knowing
Article 111 softens the edges, and most timeline summaries skip it. Three carve-outs matter in practice.
High-risk systems already on the market
A high-risk system placed on the market before 2 August 2026 is caught only once its design changes significantly after that date. A meaningful redesign pulls it straight into scope, so treat this as a pause rather than an exemption.
Public sector systems
High-risk systems already in service with public authorities get until 2 August 2030, which recognises how long government procurement and replacement cycles run.
General-purpose models
Models placed on the market before 2 August 2025 have until 2 August 2027 to meet the Chapter V obligations in full.
Related EU AI Act topics
Official references
How Prismor helps with the EU AI Act
Most of the Act's operational load comes down to two questions: which AI systems do you run, and can you show what they did. Prismor answers both for the AI inside your codebase.
Find every AI system you run
Phasing out Article 5 practices and classifying under Annex III both start from an inventory. Prismor finds the SDKs, models, agent configs, and MCP servers in your repos, including the ones nobody wrote down.
Records that hold up in an audit
Article 12 requires high-risk systems to log their operation automatically. Prismor captures every tool call and file touch with the session context an auditor asks about.
Human oversight you can enforce
Article 14 expects a person who can step in. Prismor holds risky agent actions for approval and blocks the ones that break your policy.
The AI Posture Analyzer is free. Public repos scan instantly with no signup. Private repos sign in with GitHub and run on your own access.