🤖 EU AI & Sovereignty Weekly
July 25, 2026 — 8 days until the AI Act high-risk deadline
This week’s EU AI & Sovereignty update covers the final countdown to the August 2 high-risk deadline, the Digital Omnibus delay, the new Cybersecurity & AI Action Plan, and Europe’s €37 billion bet on AI Gigafactories.
wetgeving
🔥 HOOG
EU AI Act High-Risk Rules Go Live August 2
In just 8 days, the high-risk AI obligations under the EU AI Act become enforceable — covering employment, education, critical infrastructure, law enforcement, migration, and credit scoring. Fines can reach €35 million or 7% of global annual turnover, exceeding GDPR’s maximum penalties. The AI Office gains full enforcement powers including testing protocols for GPAI models. This is the moment every European AI deployer has been preparing for — or should have been.
wetgeving
🔥 HOOG
Digital Omnibus Final: Annex III Delayed to December 2027
The EU Council formally adopted the Digital Omnibus on June 29, 2026. Standalone high-risk AI systems from Annex III (biometrics, employment, education, critical infrastructure) are pushed from August 2, 2026 to December 2, 2027 — a 16-month reprieve. AI systems embedded in regulated products (Annex I) move to August 2, 2028. Article 4 on AI literacy was softened to a best-effort obligation. The delay gives businesses breathing room, but the clock is still ticking for the core obligations.
Commissie
🔥 HOOG
EU Action Plan on Cybersecurity & AI — July 7, 2026
The European Commission published a comprehensive action plan structured around three pillars: safe use of advanced AI, strengthening cyber resilience against AI-driven threats, and building European AI leadership in cybersecurity. Key deliverables include a formal EU evaluation capacity for advanced AI models (operational in 2027) and a secure AI testing platform (end of 2026). This is the first integrated EU strategy that explicitly connects AI capability evaluation with cybersecurity.
investeringen
🔥 HOOG
Tech Sovereignty Package — June 3, 2026
The Commission adopted an ambitious package to strengthen EU digital autonomy, covering the full tech stack: semiconductors, AI, cloud, and open source. The goal is to make Europe an “AI Continent” — a world leader in AI research, development, and adoption. The package includes the €109 billion French AI plan and spans everything from chips and infrastructure to software and cloud services. This is the most comprehensive sovereignty initiative the EU has ever launched.
technology
📌 MEDIUM
19 AI Factories Operational Across Europe
The Commission confirmed 19 AI Factories are live within the EuroHPC supercomputing network, supported by 13 regional AI Factory antennas providing access for researchers, startups, and industry. The AI Gigafactories — national supercomputer clusters upgraded for frontier AI training — represent an estimated €37 billion investment. Finland, Italy, Spain, and the “Power Triangle” (NL-BE-DE) are leading the charge. This is Europe building its own AI infrastructure, piece by piece.
lidstaten
📌 MEDIUM
European AI Champions: Mistral Survives, Aleph Alpha Pivots
A deep analysis by Deluair Consultancy reveals Europe went from two AI champions to “one and a half.” Mistral AI (Paris) raised €600M Series B in June 2024 at a $6B valuation with Nvidia, Samsung, IBM, Salesforce, BNP Paribas, and Cisco. Aleph Alpha (Heidelberg) announced in November 2024 it would stop training foundation models and focus on enterprise services. DeepSeek R1’s launch in January 2025 (at $0.55/M input tokens) collapsed the cost basis for European labs in a single weekend. The lesson: European AI needs more than ambition — it needs sustainable economics.
wetgeving
🔥 HOOG
GPAI Code of Practice Enforcement Begins August 2
The General-Purpose AI Code of Practice, finalized in 2025 by the AI Office with nearly 1,000 stakeholders, becomes enforceable on August 2, 2026. GPAI providers face fines of up to 3% of global turnover or €15 million. The Code splits the model market into signatories and non-signatories — a signal now used as a procurement criterion. Models like GPT-5, Claude, Gemini, Llama 4, Mistral Large 2, and DeepSeek V3 all fall within the systemic risk threshold (10²⁵ FLOPs).
lidstaten
📌 MEDIUM
NYT: The Quest for Technological Sovereignty in Europe
The New York Times reported on July 16 on France and Germany’s struggle to become independent from US and Chinese technology. The article describes how European countries must choose where to deploy their limited resources — and that the ambition of technological sovereignty often collides with the reality of American hyperscalers (AWS, Azure, Google Cloud) dominating the European cloud market. A sobering read that highlights the gap between political ambition and market reality.
wetgeving
🔥 HOOG
38% of High-Risk AI Deployers Missing Risk Assessments
A compliance-kit.eu analysis reveals 38% of high-risk AI deployers still lack documented risk assessments under Article 9 or Article 27 of the AI Act — the largest compliance gap just three months before the August 2 deadline. Even more striking: 78% of companies haven’t started preparations at all. This mirrors the early GDPR compliance struggles, but with higher stakes and steeper fines. The August 2 deadline will be a rude awakening for many.
investeringen
🔥 HOOG
EU AI Gigafactories: The €37 Billion Sovereignty Bet
The EU is investing in AI Gigafactories — national supercomputer clusters upgraded to train frontier AI models, infrastructure that was until recently the exclusive domain of US hyperscalers and Chinese state programs. The Gigafactories are also designed to stimulate domestic production of EU AI processors as part of a “full-stack technological sovereignty” strategy. 76 Gigafactory proposals (EOIs) have been submitted. This is Europe’s biggest infrastructure bet since the Galileo satellite system.
📈 Key Trends
- • Countdown to enforcement: With 8 days until the high-risk deadline, the gap between regulatory ambition and business readiness is stark — 78% of companies haven’t started. Expect a scramble.
- • Infrastructure sovereignty: The €37 billion Gigafactory program and 19 operational AI Factories show Europe is serious about building its own compute capacity, but it’s a long-term play.
- • Champion economics: Mistral’s survival vs Aleph Alpha’s pivot illustrates the brutal economics of foundation model training — DeepSeek’s cost disruption changed the game for everyone.
- • Cybersecurity-AI convergence: The July 7 Action Plan marks the first integrated EU strategy linking AI evaluation with cybersecurity — expect more cross-domain regulation.
- • GPAI market split: The Code of Practice is creating a de facto two-tier market — signatories vs non-signatories — with procurement implications for enterprises.
Wat vond je van dit bericht?
