Use the phase dates to sequence evidence, owners, and build work.
This roadmap focuses on what should be operational in each phase, not only what the law says.
Structured answer sets in this page tree.
Cited legal and guidance references.
A strong AI Act roadmap should reduce surprise. That means using the legal phasing to bring the right controls live in the right order: early prohibitions and literacy first, GPAI documentation next, and full high risk and transparency operating controls by the main application date.
Start with inventory, roles, and classification. Without a clean register and a role matrix, every downstream workstream becomes slow and error prone. Foundation work also includes the evidence repository, the approval forum, and the first supplier documentation requests.
This phase should start as early as possible and should not wait for the first major legal date.
The first hard operating gates are Article 5 and AI literacy. Product teams, operations teams, and oversight roles should all understand what they are screening for and where escalation is mandatory. This phase also sets the tone for authority ready evidence later on.
If the foundation exists but the Article 5 gate does not, the program still has a dangerous hole.
This phase prepares for the 2 August 2025 Chapter V start date. Model providers need their Article 53 workflow and any systemic risk path. Integrators need contract and evidence leverage over upstream model suppliers.
Do not treat GPAI work as relevant only to frontier model labs. Many product teams rely on upstream models and still need supplier documentation, version notices, and incident support.
This is the deepest implementation phase because it reaches engineering design, testing, deployment, and operations. High risk systems need Articles 9 to 15 evidence, FRIA where required, and post market readiness. Transparency duties need actual product components, machine readable marking logic, QA evidence, and release review discipline.
By the time this phase ends, the organization should be able to show how the law is implemented in live systems, not only in policy documents.
After the main application date, the roadmap becomes a portfolio management problem. Legacy GPAI models, public authority deployments, and Annex X transition cases need explicit tracking. At the same time, all production systems need recurring review because the Act is sensitive to changes in intended purpose, risk profile, and system design.
This is also where the program should mature into regular audit, management review, and corrective action loops.
Assessment Autopilot can take EU AI Act (Regulation (EU) 2024/1689) Timeline and roadmap from reusing this material inside a governed evidence system to a reusable workflow inside Sorena. Teams working on EU AI Act (Regulation (EU) 2024/1689) can keep owners, evidence, and next steps aligned without copying this guide into separate documents.
Start from EU AI Act (Regulation (EU) 2024/1689) Timeline and roadmap and turn the guidance into owned tasks, evidence requests, and review checkpoints.
Review your current process, evidence gaps, and next steps for EU AI Act (Regulation (EU) 2024/1689) Timeline and roadmap.