EU AI Act Compliance Roadmap for Businesses: Deadlines and Priorities in 2026

The EU AI Act does not have a single compliance date. It follows a phased calendar with milestones that have already passed and others arriving in the coming months. Knowing exactly where your organisation stands — and what to do before which deadline — is what separates an efficient compliance strategy from a last-minute scramble.

This guide is practical: which deadlines are already enforceable, which arrive in 2026-2027, and how to structure the work based on the type of AI system you develop or use.

Deadlines already passed: what is enforceable right now

If your company uses or integrates general-purpose models to offer as a service or embed in products, the August 2025 obligations already apply: technical documentation of the model, acceptable use policy and training data summary.

August 2026: the most critical deadline for most businesses

August 2026 activates requirements for high-risk systems. This is the most relevant deadline for companies developing or using AI in HR, finance, healthcare, education or critical infrastructure.

• Risk management system: identification, analysis and mitigation of risks throughout the system lifecycle
• Data governance: documented practices for quality, representativeness and bias in training, validation and test data
• Technical documentation: full description of the system, its purpose, data used, performance metrics and known limitations
• Event logging: automatic logs enabling reconstruction of the system’s behaviour
• Transparency to the deployer: clear instructions for use, system limitations and maintenance conditions
• Human oversight: technical mechanisms enabling people using the system to intervene, pause it or reject its outputs

August 2027: legacy systems already in use

High-risk systems already on the market before the AI Act enters full application have until August 2027 to comply, unless they undergo significant design or purpose changes. This gives legacy products extra time, but does not remove the obligation.

How to prioritise: action map by company type

Five steps to structure your compliance programme

• 1. Inventory: list all AI systems your company develops, uses or contracts, including third-party models embedded in your products
• 2. Classification: determine the risk category for each system. Most will be minimal or limited risk
• 3. Gap analysis: for high-risk systems, identify which requirements are not yet met
• 4. Remediation plan: prioritise by criticality and deadline, assign owners and budget
• 5. Continuous documentation: the AI Act is not a one-off project — logging and risk management requirements are ongoing

What does AI Act compliance cost?

It depends on how many high-risk systems you have and their complexity. For a company with one high-risk system already in production, compliance costs — documentation, technical adjustments, initial audit — can range from £30,000 to £130,000, depending on whether the system was designed with these requirements in mind or needs retrofitting.

Systems designed correctly from the start — with logging, documentation and human oversight built into the architecture — are significantly cheaper to certify than those requiring retrospective adaptation.

How Yeeply can help

At Yeeply we work with certified software development companies that already incorporate AI Act requirements into their design and documentation processes. Whether you have an existing system that needs adapting or are starting a new development, we can help you get it right from the start.

Use the Request a quote button at yeeply.com/en or write to sales@yeeply.com.