On 2 December 2027, the EU AI Act rules for high-risk AI systems listed in Annex III are scheduled to apply under the current amended timeline, as amended by Regulation (EU) 2026/1744. Annex III covers specified uses in sensitive areas such as biometrics, critical infrastructure, education, employment, essential services, law enforcement, migration and border management, and justice-related contexts. The date is important because providers and deployers of systems that meet the high-risk criteria must move from preparation into the applicable high-risk compliance framework.
Why 2 December 2027 Matters
The Annex III date was moved to 2 December 2027 under the amended AI Act timeline in Regulation (EU) 2026/1744, which amended Article 113. That extension does not mean every system used in a listed sector is automatically high-risk. Classification still depends on the system's intended purpose, the specific Annex III use case and the conditions in Article 6. Organisations should therefore avoid treating labels such as 'HR AI', 'education AI' or 'biometric AI' as the final legal conclusion. The practical task is to map the real function of the system to the relevant legal wording and preserve the reasoning used to reach the classification.
Which Types of Uses Appear in Annex III?
Annex III identifies specified use cases across several sensitive areas, including certain biometric applications, critical infrastructure, education and vocational training, employment and worker management, access to essential private and public services, law enforcement, migration and border control, and administration of justice and democratic processes. The exact wording matters. Two systems used by the same organisation can receive different treatment because they perform different functions. A university chatbot, for example, should not be classified solely because it is used in education; the intended task must be tested against the relevant Annex III category and Article 6 criteria.
For the broader regulatory framework, see EU AI Act timeline.
Annex III Listing Is Not Always the End of Classification
Article 6 also matters because Annex III listing is not always the end of the analysis. The Act provides a limited route under which some listed systems may not be treated as high-risk when they do not pose a significant risk of harm and perform specified narrow tasks, subject to the conditions in the provision. Profiling of natural persons changes that analysis. Organisations relying on an exception should document why the conditions are met, what the system actually does and what evidence supports the conclusion. A marketing description or a supplier's informal assurance is not a substitute for the classification record.
What High-Risk Providers Need to Have in Place
For providers of high-risk systems, the applicable framework includes requirements across risk management, data and data governance where relevant, technical documentation, record-keeping, information for deployers, human oversight, accuracy, robustness and cybersecurity. Provider duties also include the appropriate conformity, registration, quality-management and post-market processes where the Act requires them. These controls should connect to the system's intended purpose and technical design. A provider should be able to trace each relevant requirement to design decisions, test evidence, instructions, release records and the people responsible for maintaining the system after it reaches the market.
What High-Risk Deployers Need to Do
Deployers of high-risk systems have a different role. Their obligations focus on operating the system according to instructions, assigning appropriate human oversight, monitoring use, handling input data where relevant and retaining logs under the conditions of Article 26. Certain deployments can carry additional duties, including information or impact-assessment requirements. A buyer should therefore not assume the provider's conformity work completes the organisation's own compliance. The deployment context, affected people, staff responsibilities, escalation process and actual use of outputs need their own documented review and operational controls.
Continue with what changes on 2 December 2026.
What Evidence Should Be Ready by the Application Date?
Enforcement of Annex III high-risk obligations follows the date on which those provisions become applicable. By 2 December 2027, organisations should already have completed classification, ownership and evidence work rather than beginning discovery on the deadline itself. Useful records include the system version, intended purpose, Annex III category considered, Article 6 reasoning, provider and deployer identities, technical documentation, oversight arrangements, testing evidence and applicable registrations or assessments. Changes to purpose, integration, model behaviour or user population should trigger reassessment because the original classification and control set may no longer fit the system.
Annex I Product-Related High-Risk Rules Still Come Later
The 2 December 2027 milestone does not cover every high-risk route. AI systems that are high-risk because they are embedded in regulated products covered by Annex I have a later application date of 2 August 2028 under the current amended timeline in Regulation (EU) 2026/1744. Separately, Member States should have at least one AI regulatory sandbox operational by 2 August 2027, in the Service Desk wording, which is a governance milestone that precedes the Annex III date. Organisations should therefore identify whether a candidate system follows the Annex III use-case route or the Annex I product route before assigning a deadline. Keeping the routes separate prevents premature assumptions and makes implementation planning clearer, especially for companies that sell both standalone software and AI-enabled regulated products across multiple markets.
Frequently asked questions
Do all AI systems used in Annex III sectors become high-risk on 2 December 2027?
No. The specific intended purpose, listed use case and Article 6 conditions must be assessed. Sector presence alone does not determine classification, and the Article 6(3) route and profiling limit must be tested where relevant.
What is the main difference between the Annex III and Annex I dates?
Annex III covers specified high-risk use cases and applies from 2 December 2027, while the high-risk rules for AI embedded in regulated products covered by Annex I apply from 2 August 2028, both as amended by Regulation (EU) 2026/1744.
Should providers wait until December 2027 to prepare?
No. Classification, technical documentation, risk management, testing, oversight design and deployment planning can require substantial lead time and should be organised before the application date, with sandbox and guidance milestones tracked in the meantime.
What does this article not decide?
This article does not decide whether any particular system is high-risk or which conformity route it must follow. Those questions turn on the system's intended purpose, the Annex route and the Article 6 conditions, and should be confirmed with a competent authority or legal adviser where they are contested.
Continue through this AI Act cluster
This article is an informational overview, not a legal determination. Check the current legal text and official implementation guidance for your system and role.
Consolidated Regulation (EU) 2024/1689, as amended ↗AI Act Service Desk implementation timeline ↗AI Act Service Desk: Article 6 ↗