
🤖 EU AI & Sovereignty Weekly
5 July 2026
This week’s EU AI & Sovereignty update covers the final approval of the Digital Omnibus, the landmark Tech Sovereignty Package, and the fast-approaching GPAI enforcement deadline of 2 August 2026.
wetgeving
🔥 HOOG
EU Council Gives Final Green Light to Digital Omnibus on AI
The Council of the EU gave its final approval on 29 June 2026 to the Digital Omnibus on AI, part of the Omnibus VII package. This formalises the postponement of high-risk AI obligations: Annex III systems (employment, education, biometrics, critical infrastructure) are shifted to 2 December 2027, and AI systems embedded in regulated products to 2 August 2028. The European Parliament had already approved on 16 June. The law is now final, marking the first major amendment to the AI Act since its adoption in June 2024.
Commissie
🔥 HOOG
Tech Sovereignty Package: Four Laws for Digital Autonomy
The European Commission presented the European Technological Sovereignty Package on 3 June 2026, consisting of four components: (1) Chips Act 2.0, (2) the Cloud and AI Development Act (CADA), (3) an EU Open Source Strategy, and (4) a strategic roadmap for AI in the energy sector. The EU depends on non-EU suppliers for more than 80% of its digital products and services — a vulnerability Brussels now treats as a strategic risk comparable to its former gas dependency on Russia.
wetgeving
🔥 HOOG
Cloud and AI Development Act (CADA): Four Sovereignty Levels
CADA forms the centrepiece of the Tech Sovereignty Package, introducing a four-tier sovereignty framework for cloud procurement. Member states must assess jurisdiction risks for digital government contracts. At the highest level (critical government systems), US hyperscalers (AWS, Azure, Google Cloud) are excluded unless they can demonstrably guarantee European law compliance — targeting the US CLOUD Act. The proposal also includes €20 billion in EU data centre and AI infrastructure investments.
wetgeving
📌 MEDIUM
Digital Omnibus Agreement (May): High-Risk Deadlines Revised
On 7 May 2026, the Council and Parliament reached a provisional political agreement on the Digital Omnibus, the first set of amendments to the AI Act since its adoption. Besides postponing high-risk deadlines, the package includes targeted simplifications and new prohibitions. GPAI obligations and the fine framework (up to 3% of global turnover for GPAI violations) remain in force from 2 August 2026. System-risk AI model obligations have been applicable since August 2025.
technology
🔥 HOOG
GPAI Enforcement Starts 2 August 2026 — 28 Days Left
With 28 days remaining, full enforcement for General-Purpose AI (GPAI) providers kicks in. Fines up to 3% of global annual turnover or €15 million become possible. The AI Office has prepared a Code of Practice that providers can follow as a ‘safe harbour’. The Code effectively splits the model market into signatories and non-signatories — a distinction that is already becoming a procurement signal. Transparency obligations (Art. 50) and the fine structure also activate on 2 August.
investeringen
🔥 HOOG
35 New NVIDIA AI Supercomputers in Development Across Europe
NVIDIA announced at ISC High Performance 2026 that a record 35 new AI supercomputers are under development across 23 European countries. These include systems at the Barcelona Supercomputing Center, BavariaAI’s Blue Swan, IT4LIA, HLRS’s HammerHAI, and NAISS’s Mimer EuroHPC AI Factory. The systems will provide over 3 million researchers with next-generation AI infrastructure — Europe’s largest single-year expansion of supercomputing capacity.
investeringen
🔥 HOOG
EuroHPC AI Factories: €1.5 Billion for Seven AI Factories
EuroHPC JU has allocated €1.5 billion for the first AI Factories in Europe. Seven locations have been selected: Finland, Germany, Italy, Luxembourg, Sweden, and two others. Five new AI-optimised supercomputers will be operational by the end of 2026. The AI Factories aim to more than double EuroHPC’s computing power, giving European startups and researchers access to the compute needed to train advanced AI models on home soil.
lidstaten
📌 MEDIUM
European AI Startups: Record $8.4 Billion Investment in Q1 2026
European AI startups raised $8.4 billion in Q1 2026, up nearly 30% year-on-year. AI accounted for more than 50% of total venture capital in Europe for the first time. The four largest funding rounds in Q1 2026 all went to AI companies. Mistral AI (France) remains Europe’s leader with a valuation of ~€11.7 billion, followed by Aleph Alpha (Germany, sovereign AI pioneer), Wayve (UK, autonomous driving), Nscale (Norway, AI infrastructure), and ElevenLabs (UK, AI voice).
wetgeving
📌 MEDIUM
GPAI Code of Practice — Final Framework for Foundation Models
The final GPAI Code of Practice has been drafted by a group of independent experts in a multi-stakeholder process. It provides a voluntary framework for general-purpose AI providers to demonstrate AI Act compliance. The Code covers transparency documentation, copyright policy, risk management (for systemic risk models), model evaluations, incident reporting, and cybersecurity. Providers following the Code receive a presumption of conformity. Final details are being wrapped up in July 2026, just ahead of the 2 August enforcement deadline.
wetgeving
📌 MEDIUM
Chips Act 2.0: Emergency Powers for Chip Crises
The Chips Act 2.0 grants the European Commission emergency powers to intervene in commercial contracts of chip manufacturers during shortages. Brussels can force companies to prioritise EU crisis orders over existing contracts. Companies withholding supply chain information during a shortage risk fines up to €300,000. A joint chip purchasing mechanism — modelled on the EU vaccine procurement during the pandemic — aims to give smaller member states more bargaining power against major foundries. The original Chips Act from 2023 targeted 20% of global chip production by 2030, but that goal is under pressure.
📈 Key Trends
- • The EU is shifting from regulation-only to active investment: the Tech Sovereignty Package (€20B+ in cloud/AI infra) combined with EuroHPC AI Factories (€1.5B) signals a new phase of European digital industrial policy.
- • The US CLOUD Act is now a named policy target — CADA explicitly excludes US hyperscalers from critical government contracts, marking a major escalation in digital sovereignty.
- • GPAI enforcement starting 2 August creates a compliance cliff: the Code of Practice split (signatories vs non-signatories) is already affecting procurement decisions.
- • European AI startups are flourishing: $8.4B in Q1 with AI taking >50% of all VC for the first time — Mistral, Aleph Alpha, and Wayve are leading a genuine European AI ecosystem.
- • Infrastructure buildout is accelerating: 35 new NVIDIA supercomputers across 23 countries plus 7 EuroHPC AI Factories will more than double Europe’s AI compute capacity.
Wat vond je van dit bericht?