🤖 EU AI & Sovereignty Weekly
July 14, 2026
This week’s EU AI & Sovereignty update covers the Cloud and AI Development Act (CADA), the EUROPA consortium’s 400B+ open model, EuroHPC AI Factories, the Fable 5 export ban fallout, and more. Europe’s push for digital sovereignty is accelerating at an unprecedented pace.
wetgeving
🔥 HIGH
EU Presents Cloud and AI Development Act (CADA)
The European Commission unveiled the Technology Sovereignty Package on June 3, with CADA as its centerpiece. It introduces four assurance levels for cloud services, requiring public institutions to conduct risk assessments before procurement. The act mandates sovereignty risk assessments for banks, healthcare, and energy sectors, aiming to triple European computing power within 5–7 years.
technology
🔥 HIGH
EUROPA Consortium Selected for 400B+ Open AI Model
The European Commission selected the EUROPA consortium, led by Italy’s Domyn, to build a 400+ billion parameter open-source AI model covering all 24 EU languages. Funded through the Frontier AI Grand Challenge and trained on EuroHPC infrastructure, the project is expected to be delivered in 12–18 months (end of 2027). This is the first EU initiative aimed at being frontier-competitive rather than just “good for a European model.”
investeringen
🔥 HIGH
€1.5 Billion EuroHPC AI Factories: Seven Supercomputers
EuroHPC has allocated €1.5 billion for the first seven AI Factories across Finland, Germany, Greece, Italy, Luxembourg, Spain, and Sweden. With five new AI-optimized supercomputers expected operational by end of 2026, and a record 35 new NVIDIA AI supercomputers at European research institutions, the investment will more than double EuroHPC’s computing capacity.
technology
📌 MEDIUM
Apertus 70B: Best Open European AI Model Available Now
Switzerland’s Apertus (EPFL, ETH Zürich, CSCS, Swisscom) is the strongest open European AI model available today. It comes in 8B and 70B parameter variants under Apache 2.0 license, trained on 15 trillion tokens across 1,811 languages. While not frontier-competitive with GPT-5.5 or GLM-5.2, it significantly outperforms American models on European languages like Swiss German and Romansh.
lidstaten
📌 MEDIUM
Mistral AI: €13.8B Valuation, $400M ARR
French AI company Mistral AI raised a $2B Series C at a $13.8 billion valuation, with $400M annual recurring revenue. It remains the only European AI company with commercial frontier models. Since the Fable 5 export ban, Mistral has sharpened its sovereignty narrative — the Anthropic shutdown handed Mistral its sovereignty argument on a silver platter. Caveats: larger models are not fully open-weight, and there is non-European investment.
Commissie
🔥 HIGH
Fable 5 Export Ban Accelerates European AI Sovereignty
The Fable 5 export ban (June 12, 2026) fundamentally changed the European AI strategy. Before the ban, “sovereign AI” was a nice-to-have; afterwards, European companies lost access to their primary AI tools (OpenAI, Anthropic, Google) overnight. The Atlantic Council report states the Trump administration inadvertently drove the EU toward technological independence. Procurement policies are now being rewritten to require proof that suppliers cannot be cut off.
wetgeving
📌 MEDIUM
New Wave of EU AI Act Obligations Takes Effect
The AI Act (Regulation 2024/1689) — the world’s first comprehensive AI legal framework — has entered a new wave of obligations. High-risk AI systems in banking, insurance, and HR must now be auditable, transparent, and robust. General-purpose AI model obligations (transparency, documentation, copyright) are phased in through 2027–2028. The EUROPA model is being designed AI Act-compliant from the start.
lidstaten
📌 MEDIUM
Berlin Summit: France & Germany Agree on Data Governance
The Summit on European Digital Sovereignty in Berlin (November 2025) delivered landmark commitments. France and Germany pushed for a unified, resilient EU data governance architecture that protects sensitive data while supporting economic competitiveness. A joint declaration established strong common positions on AI Gigafactory (AIGF) selection criteria and EuroHPC regulation governance. This event laid the political groundwork for the June 2026 Tech Sovereignty Package.
Commissie
📌 MEDIUM
EU Open-Source Strategy: ‘Open Source First’ in Law
The Tech Sovereignty Package includes an EU Open-Source Strategy that enshrines “open source first” as an operational principle in law. This provides significant tailwind for bring-your-own-model architectures and European cloud providers offering open-source AI models. The principle ensures public funds contribute to reusable, transparent technology rather than vendor lock-in, and is one of four pillars alongside CADA, Chips Act 2.0, and the AI energy roadmap.
wetgeving
📌 MEDIUM
CRA Article 14: 24-Hour Cyber Mandate from September 11
The Cyber Resilience Act (CRA) Article 14 takes effect on September 11, 2026, requiring manufacturers and providers of digital products to report actively exploited cyber incidents within 24 hours that pose a cybersecurity risk. This applies to AI models and platforms sold in the EU, directly impacting AI infrastructure and services. It represents a significant compliance requirement for the entire European AI ecosystem.
📈 Key Trends
- • CADA is the biggest game-changer: digital sovereignty is now legally enforceable in public procurement, with four certified assurance levels
- • EUROPA must bridge the gap: Europe has no frontier-competitive open model yet — Apertus 70B is the best available but doesn’t match GPT-5.5 or GLM-5.2
- • Fable 5 ban as catalyst: the US export ban transformed sovereignty from “nice-to-have” into an urgent business requirement
- • Infrastructure investment is accelerating: €1.5B for AI Factories plus 35 new NVIDIA supercomputers signal serious commitment to European compute capacity
- • Regulatory pressure is mounting: AI Act obligations, CRA cyber mandates, and open-source-first policies are creating a complex but coherent compliance landscape
Wat vond je van dit bericht?
