The deadline extension bought time. These guidelines tell you what to do with it.
According to IAPP reporting, the European Commission has released draft guidelines on how to determine whether an AI system qualifies as high-risk under Article 6 of the EU AI Act. The timing is deliberate: Annex III’s compliance deadline just moved to December 2, 2027, giving organizations 16 more months. These guidelines tell them what they’re actually working toward.
The draft uses a three-phased approach to making the Article 6 high-risk determination, per IAPP. It provides non-exhaustive examples for Annex I and Annex III use cases, the two pathways through which systems can be classified as high-risk. According to DataGuidance reporting, the guidelines also clarify the conditions under which Annex III systems may qualify for exemption from high-risk classification under Article 6(3). That exemption pathway matters: a system that meets the surface criteria for Annex III may still fall outside high-risk obligations if it satisfies specific conditions the guidelines are now defining.
Unanswered Questions
- Does the three-phased classification approach align with how your current AI inventory and risk taxonomy is structured?
- Does your AI system meet the Article 6(3) exemption conditions the guidelines now define, and have you assessed this formally?
- Does your organization plan to submit a response to the consultation before June 23?
Article 6 Consultation Action Checklist
- Confirm June 23 close date on Commission's official consultation page
- Review draft guidelines' three-phased determination framework against current AI inventory classification
- Assess Article 6(3) exemption eligibility for any Annex III candidate systems
- Decide whether to submit consultation response before June 23 deadline
The consultation period is reported to close June 23, 2026, according to IAPP. That date hasn’t been independently verified beyond single-source reporting, treat it as the operative planning date while confirming directly with the Commission’s official consultation page.
The catch: Article 6 classification is the gateway determination. Get it wrong and your compliance program is built on a false foundation, either you over-invest in a program you don’t need, or you under-invest and miss obligations you do. The Commission delayed these guidelines past a February 2026 target. Their arrival now, post-Omnibus, is the signal that the classification infrastructure is catching up to the deadline structure.
The three-phased approach itself is new framing. The EU AI Act provides Article 6(1)(a) and 6(1)(b) as the dual classification pathways, product safety integration and standalone use case enumeration, but the draft guidance adds a structured determination process that goes beyond the statutory text. That’s what practitioners needed: not just what the law says, but how the Commission expects organizations to work through the question.
Definition
Don’t expect this guidance to close every gap. Article 6(3) exemption clarity helps, but harmonized technical standards for conformity assessment haven’t published. Classification guidance and conformity standards are different documents serving different steps in the same workflow. The guidance tells you whether you’re in. The standards tell you what “in” requires you to build.
The real question for compliance teams: does the three-phased framework the guidelines introduce align with how your internal AI inventory and risk classification process is structured? If your organization built its classification methodology before this guidance dropped, the June 23 consultation window is the opportunity to flag mismatches and submit input before the final version locks. After June 23, you’re working with whatever the Commission finalizes.