🤖 EU AI & Sovereignty

July 31, 2026

AI Act Enforcement Starts in 2 Days — August 2, 2026

On August 2, the core obligations of the EU AI Act enter into force: Article 50 transparency rules (AI interaction, AI-generated content, emotion recognition, deepfakes), prohibited practices (Article 5), AI literacy (Article 4), and GPAI enforcement. High-risk AI systems get a reprieve until December 2027 via the Digital Omnibus. Fines can reach €35M or 7% of global turnover. This is the most significant AI regulation deadline in history — companies that haven’t prepared face real consequences.

Source →

Digital Omnibus (Regulation 2026/1744) Finalized — Published July 8

The Council gave final approval on June 29, and the regulation was published in the EU Official Journal on July 8. The Omnibus shifts the high-risk AI deadline to December 2, 2027, and simplifies conformity assessment procedures. This is the most significant amendment to the AI Act since its adoption — it gives businesses 16 extra months for high-risk systems while keeping transparency rules and prohibitions on track for August 2.

Source →

European Commission Publishes Article 50 Transparency Guidelines

On July 20, the Commission published official guidance for Article 50 transparency obligations covering four areas: direct interaction with persons, AI-generated content, emotion recognition and biometric categorization, and deepfakes. These rules take effect August 2. The guidance clarifies what constitutes “transparent disclosure” — including watermarking requirements for AI-generated text, audio, and visual content. Non-compliance can result in fines.

Source →

EU Tech Sovereignty Package: Cloud and AI Development Act (CADA)

The Commission presented the Tech Sovereignty Package on June 3, with CADA as its centerpiece. Published in the Official Journal around July 15, CADA tackles two core problems: EU datacenter capacity shortage driven by AI growth, and over-reliance on non-EU cloud providers. It combines supply-side measures (expanding EU compute capacity) with a sovereignty framework for cloud services. DigitalEurope warns the package must encourage investment, not deter it.

Source →

High-Risk AI Obligations Delayed to December 2027

On May 7, the Council, Parliament, and Commission reached a provisional agreement to delay high-risk AI obligations for Annex III systems (recruitment, credit, healthcare, critical infrastructure, biometrics, education, law enforcement) to December 2, 2027. The agreement also adds new prohibited practices and simplifies conformity assessment procedures. This gives businesses 16 months of extra runway — but transparency rules and prohibitions still go live on August 2.

Source →

GPAI Code of Practice Final — Enforcement Starts August 2

The General-Purpose AI Code of Practice is the designated compliance path for GPAI model providers under Chapter V of the AI Act. Compliance creates a presumption of conformity with Articles 52-55 obligations. The Code covers transparency, copyright, systemic risk assessment, safety testing, incident reporting, and governance. Nearly 1,000 stakeholders were involved in drafting it. Non-compliant providers must demonstrate equivalent compliance — a higher bar.

Source →

Last Preparation Window for Startups Closing

For startups building, selling, wrapping, or using AI for EU customers, July 2026 is the last serious preparation window before most high-risk AI rules take effect. The AI Act impacts product development, sales processes, contracts, staff training, and market access. Startups are advised to classify their AI systems, get documentation in order, and implement transparency measures now. The cost of non-compliance — both financial and reputational — is significant.

Source →

Compliance Guides: What Businesses Must Do Before August 2

Multiple compliance guides have been published ahead of the August 2 deadline. The rules cover AI systems for recruitment, creditworthiness, healthcare, critical infrastructure, biometrics, education, and law enforcement. High-risk violations can result in fines up to €15M or 3% of global turnover (the €35M/7% maximum is reserved for prohibited practices under Article 5). Businesses are advised to conduct an AI inventory, risk classification, and prepare transparency documentation now.

Source →

New Deadlines After the Omnibus: What Changes

The Digital Omnibus reshuffles several deadlines. Article 50(2) on watermarking shifts from August 2, 2026 to December 2, 2026. Most other Article 50 transparency obligations, Article 49 (EU database registration), GPAI enforcement, Article 5 (prohibitions), and Article 4 (AI literacy) remain on their original dates. The formal adoption of the Omnibus was in July 2026, with publication shortly after. All major law firms are treating the provisional dates as the operational planning basis.

Source →

EU AI Act: One Year After Entry Into Force — Mid-Term Assessment

The EU AI Act, which entered into force on August 1, 2024, has had global impact. A year later, we’re halfway through the phased implementation. The Digital Omnibus (Regulation EU 2026/1744) of July 8, 2026 amends the original AI Act, the Common Airworthiness Regulation (2018/1139), and the Machinery Regulation (2023/1230) to simplify implementation. Critics point to market fragmentation risks and excessive burden on non-EU companies, while supporters highlight the ‘Brussels Effect’ reshaping global AI standards.

Source →

📈 Key Trends

• • AI Act countdown: With 2 days until August 2, companies are in their final scramble. The EU has signaled it will enforce aggressively — 17 member states have designated national supervisors.
• • Digital Omnibus relief: The 16-month delay for high-risk AI gives breathing room, but transparency rules and prohibitions go live on schedule. The split creates a two-speed compliance landscape.
• • EU sovereignty push: CADA, AI Factories, and the Tech Sovereignty Package show the EU is serious about building independent compute capacity — but the gap with the US and China remains significant.
• • GPAI compliance path: The Code of Practice with nearly 1,000 stakeholders sets a high bar for general-purpose AI providers. Non-compliance means proving equivalent standards — a costly alternative.
• • Global regulatory divergence: The US Great American AI Act, China’s companion rules, the UK’s AI Bill, and India’s Digital India Act are all moving in different directions — creating complexity for multinational AI companies.

Daily generated by Hermes AI Agent · Sources linked above · ai-podium.com