GLACIS·EU AI Act series·Updated August 2026
EU AI Act high-risk compliance, before the cliff edge.
The Act can apply to providers and deployers inside or outside the EU when its territorial conditions are met. Prohibited-practice fines can reach €35M or 7% of global turnover. The AI Omnibus entered into force on 27 July 2026: relevant high-risk obligations apply from 2 December 2027 for Annex III systems and 2 August 2028 for Annex I product-embedded systems. Technical documentation and Article 12 logging must be ready when the applicable duties begin.
After a provisional agreement in May 2026, the AI Omnibus was formally adopted as Regulation (EU) 2026/1744 and entered into force on 27 July 2026. Relevant Annex III high-risk obligations now apply from 2 December 2027, and relevant Annex I product-embedded obligations from 2 August 2028. Other AI Act obligations retain their own application dates; the amendment did not postpone the Act as a whole.
The dates are reflected in the consolidated AI Act on EUR-Lex. Classification and the applicable conformity pathway remain system-specific; this page is general information, not legal advice.
High-risk systems under Annex III
For an Annex III use case, Article 6(2) classification turns on the system’s intended purpose and the specific Annex III entry. Article 6(3) provides a limited route for a listed system that does not pose a significant risk of harm to health, safety, or fundamental rights, subject to its conditions; profiling systems listed in Annex III remain high-risk. A system classified high-risk is subject to Articles 9 to 15 (risk management, data governance, technical documentation, logging, transparency, human oversight, accuracy/robustness/cybersecurity), plus Article 17 quality management.
| Domain | Typical systems in scope |
|---|---|
| Biometrics | Remote identification, categorization, emotion recognition (outside law-enforcement carve-outs) |
| Critical infrastructure | Safety components for water, gas, electricity, traffic management, digital networks |
| Education & vocational training | Admissions scoring, exam evaluation, attainment-level assignment, prohibited-behavior detection |
| Employment | Recruitment, selection, performance evaluation, termination, work allocation |
| Essential services | Creditworthiness, life and health insurance pricing, public-benefit access decisions, emergency triage |
| Law enforcement | Risk assessment of natural persons, polygraphs, evidence reliability, profiling |
| Migration, asylum & border | Risk assessment, document verification, application examination support |
| Justice & democratic processes | Judicial-decision support, alternative dispute resolution, election influence systems |
Employment screening is Annex III high-risk: recruitment, ranking and evaluation systems carry the full Articles 9 to 15 load. See how signed runtime evidence works for hiring AI. Clinical AI embedded in regulated products follows the embedded-product track; start with medical devices.
What Articles 9 to 15 actually require
| Article | Requirement |
|---|---|
| Art. 9 | Risk management system across the lifecycle: identify, evaluate, mitigate, monitor. |
| Art. 10 | Data governance for training, validation, and testing, covering relevance, representativeness, and error checks. |
| Art. 11 | Technical documentation per Annex IV (nine substantive sections). |
| Art. 12 | Automatic event logging capabilities appropriate to the covered high-risk system’s intended purpose. Configured GLACIS paths can contribute scoped records for selected events. |
| Art. 13 | Transparency and instructions for downstream deployers. |
| Art. 14 | Effective human oversight measures. |
| Art. 15 | Accuracy, robustness, and cybersecurity, including resilience to adversarial input. |
| Art. 17 | Quality management system covering compliance, post-market monitoring, incident reporting. |
Penalty structure under Article 99
Fines fall into three bands, and which ceiling applies depends on the size of the undertaking. For undertakings other than SMEs, the higher fixed amount or turnover percentage is the ceiling; for SMEs, each fine must not exceed the lower of those two ceilings. National competent authorities set the actual fine within the applicable ceiling; the AI Office handles GPAI providers directly.
| Violation | Maximum fine | Or % of global turnover |
|---|---|---|
| Prohibited practices (Article 5) | €35,000,000 | 7% |
| Other non-compliance (Articles 9 to 15, 17, etc.) | €15,000,000 | 3% |
| Incorrect information to authorities | €7,500,000 | 1% |
Enforcement and institutional design remain phased across the Union. Confirm the current competent authority, applicable system category, and effective date before relying on this overview for a specific deployment.
How GLACIS fits the obligations
GLACIS can connect configured controls to signed operational records for a defined AI workflow. Those records may support an organization’s Article 12 logging and review process; they do not replace technical documentation, conformity assessment, or legal analysis.
| Article | What GLACIS produces |
|---|---|
| Art. 9 Risk management | Scoped records showing which configured controls evaluated an in-scope action and which outcome was reported. |
| Art. 11 Technical docs | Operational evidence that can supplement, but does not generate or replace, Annex IV documentation. |
| Art. 12 Logging | Signed event records with explicit scope, identity, timestamps, control outcomes, and integrity checks. |
| Art. 14 Human oversight | Records of configured escalation, review, and override events for covered actions. |
| Art. 15 Robustness | Evidence from configured tests and runtime controls, interpreted alongside system-level evaluation. |
| Art. 17 QMS | Evidence artifacts that can support an organization’s quality-management and post-market processes. |
Signed, scoped records may contribute to an Article 12 evidence set when their fields and coverage are relevant; they do not replace the system’s logging design or establish conformity.
Go deeper
| Full compliance guide | Risk categories, Articles 9 to 15 in detail, GPAI obligations, conformity assessment paths, the Omnibus status. |
| For Chief Compliance Officers | Programme architecture, audit-readiness checklist, board reporting, certification routes. |
| For CISOs | Article 12 logging architecture, Article 15 robustness, sec-eng integration. |
| For General Counsel | Liability allocation, vendor and deployer contracts, extraterritorial scope. |
| EU AI Act vs HIPAA | Crosswalk for healthcare and life-sciences operators with US obligations. |
| Colorado ADMT law (SB 26-189) | The US transparency analogue: covered automated decision-making technology, with substantive compliance from 1 January 2027, and what stacks with the EU regime. |
By member state
| Germany | Bundestag adopted KI-MIG on 11 June 2026, assigning roles to BNetzA, sector authorities and KoKIVO; verify promulgation, entry into force and the current authority route. |
| France | Decentralised model: CNIL on workplace/education emotion-recognition; ANSSI on cybersecurity; PEReN technical support; the multi-authority DDADUE bill passed the Senate on 18 February 2026 and remains before the National Assembly. |
| Italy | National AI Law No. 132/2025 in force 10 October 2025; AgID notifying authority, ACN market surveillance, Garante on data; delegated implementing decrees have a statutory deadline of 10 October 2026. |
| Spain | AESIA operational since June 2024; 16 detailed compliance guides published December 2025; regulatory sandbox; draft national AI Law (March 2025). |
| Netherlands | Proposed hybrid 10-authority model led by AP, with AP+RDI co-coordination; public consultation on the proposed Implementation Act ran from 20 April to 1 June 2026 and is closed. |
| Belgium | BIPT designated main market surveillance authority (2025-2029 Federal Government Agreement); 21 fundamental-rights bodies under Article 77. |
| Poland | New body KRiBSI under construction (single-authority model); operational support nested in Ministry of Digital Affairs; UODO disputing advisory-only role. |