EU AI Office: draft guidelines for classifying high-risk AI systems
The European Commission opened a targeted public consultation on 13 May 2026 on draft guidelines for classifying AI systems as high-risk under the EU AI Act. The consultation closes on 22 May at 18:00 CET, and the guidelines will directly determine which organisations in healthcare, education, critical infrastructure, and HR processes must meet the strictest regulatory requirements.
This article was generated using artificial intelligence from primary sources.
The European Commission opened a targeted public consultation on 13 May 2026 on draft guidelines for classifying AI systems as high-risk under the EU AI Act. The consultation closes on 22 May 2026 at 18:00 CET — giving affected organisations, industry, and civil society just over 36 hours from the time this article was published to submit formal feedback to the EU AI Office.
Why are the classification guidelines central to the AI Act?
The EU AI Act was adopted in 2024 and enters full enforcement in two phases between 2026 and 2027. The high-risk category is the most stringent tier of regulation and covers AI systems in healthcare, education, employment, critical infrastructure, justice, border control, and access to public services. Classification determines whether an AI system will be subject to the full conformity assessment regime, CE marking, registration in the EU database, and continuous post-market monitoring — or to lighter transparency obligations.
The guidelines currently under consultation are not law; they operationalise how the EU AI Office and national authorities will interpret the ambiguous boundaries between categories. The practical stakes are significant: an organisation that overestimates its risk level bears compliance costs it need not; one that underestimates it faces penalties of up to €35 million or 7% of global turnover.
Which sectors need to follow this consultation urgently?
Specifically, the guidelines will clarify borderline cases: HR tools for CV screening and internal performance review models; clinical decision support systems (radiology AI, sepsis prediction); educational platforms that carry out adaptive testing or rank students for admission; credit scoring algorithms for access to public subsidies; biometric identification in employment or service contexts.
Companies from sectors that are at least partially affected — Telia, Bayer, Siemens Healthineers, Saint-Gobain, as well as US companies importing AI products into the EU market — have 36 hours to submit feedback. Academic institutions and NGOs may also submit written opinions.
Parallel consultation on transparency obligations
The EU AI Office is simultaneously running a consultation on transparency obligation guidelines under Art. 50 of the AI Act (open since 8 May), which governs disclosure when users interact with an AI system (e.g. chatbots), as well as synthetic content markers (deepfakes, AI-generated imagery). Two fundamental implementing measures of the AI Act are thus being defined simultaneously, making legal analysis considerably more complex for companies since the two measures interact with each other.
The EU AI Office’s final guidelines are expected before 2 August 2026, when the main obligations for high-risk AI systems and general-purpose AI models enter into force.
Frequently Asked Questions
- Which sectors are affected by the high-risk AI classification?
- The AI Act defines high risk for AI systems in healthcare (diagnostics, triage), education (student assessment, admissions), employment (CV screening, performance review), critical infrastructure (energy, water, transport), justice, border control, and access to public services. The guidelines precisely define which specific use cases fall within these categories.
- What is required of organisations developing high-risk AI systems?
- Organisations must implement risk management, data governance, technical documentation, human oversight, transparency disclosure, robustness, and accuracy. Systems must pass a conformity assessment (CE marking) and be registered in the EU database before being placed on the market.
- Is there a parallel consultation?
- The European Commission is running a parallel consultation on transparency obligation guidelines for AI systems (Art. 50 of the AI Act), open since 8 May 2026. Two key implementing measures of the AI Act are thus being defined simultaneously through industry feedback.