The EU AI Act takes effect on August 2, 2026.
Most cloud AI can't meet the new high-risk rules. On-premise can — we'll show you how.
Time until enforcement
Three obligations most teams aren't ready for
Art. 6 — Risk classification
Every AI system used commercially must be classified (prohibited / high-risk / limited / minimal). HR screening, credit decisions, and medical triage default to high-risk.
Art. 13 — Transparency obligations
Users must be informed when interacting with AI. AI-generated content must be labeled. High-risk systems require technical documentation describing capabilities, limitations, and intended use.
Art. 14 — Human oversight
High-risk systems must have a designated human supervisor with override authority, trained on the system's behavior. Records of overrides must be retained.
Why cloud AI struggles under high-risk classification
On-premise eliminates three audit vectors that are deal-breakers for high-risk classification.
3-step compliance roadmap
From first audit to a defensible compliance posture — typical timeline 8–12 weeks.
Classify your AI use cases
We map every existing AI use case to its EU AI Act category (Annex III, Art. 5–6). Output: a documented risk register your DPO can sign off on.
Typical duration: 2 weeks
Migrate high-risk workloads on-prem
Provision a WerkHub appliance sized for your workload. Move RAG indexes, model selection, and inference into your network. Keep low-risk use cases on cloud if you want.
Typical duration: 4–6 weeks
Audit documentation + training
Technical documentation per Art. 11. Inference logging configured per Art. 12. Human-oversight roles and training delivered for designated supervisors.
Typical duration: 2 weeks
Choose what you need next
This page is informational and does not constitute legal advice. For binding interpretation, consult your DPO or a specialised law firm. The August 2, 2026 enforcement date applies to high-risk AI systems under Regulation (EU) 2024/1689; some provisions phase in earlier or later.