
🤖 EU AI & Sovereignty Weekly
Friday, 7 August 2026
This week’s EU AI & Sovereignty update. The AI Act is now in official enforcement for five days. This edition combines the big ongoing stories (AI Act enforcement, AI Gigafactories, Scaleup Europe Fund) with fresh items: the strong backing for the Code of Practice on AI-generated content transparency (~190 organisations), the fourth meeting of the GPAI Signatory Taskforce, the new enforcement powers of the Commission vis-à-vis frontier-AI companies (OpenAI, Anthropic), and the news that the AI Gigafactories call has drawn no fewer than 77 proposals from 16 member states.
regulation
🔥 HIGH
AI Act enforcement and transparency rules in force since 2 August
Since 2 August, the Commission’s AI Office and national authorities enforce the AI Act, and the new transparency rules apply: chatbots must disclose they are AI, deepfakes must be labelled and AI-generated content must carry machine-readable markings. The first complaints and whistleblower tools are operational, and the first concrete enforcement actions are expected in the coming weeks. This remains the defining story of the European AI legal framework this autumn.
Commission
🔥 HIGH
~190 organisations back the Code of Practice on AI-generated content transparency
Around 190 organisations from a wide range of sectors (IT, telecom, education, retail) signed the Code of Practice on transparency of AI-generated content ahead of the marking and labelling obligations taking effect on 2 August. Leading AI companies signed section 1 — including Aleph Alpha, Anthropic, Black Forest Labs, Cohere, Google, Meta, Microsoft, Mistral and OpenAI — while firms like Bulgari, Getty Images, Lenovo and Lufthansa signed section 2; roughly half of the signatories are small and recent companies. This gives providers a streamlined, legally certain route to compliance.
regulation
📌 MEDIUM
Fourth GPAI Signatory Taskforce meeting on Safety & Security and Copyright
On 17 July, the GPAI Signatory Taskforce discussed the Safety & Security and Copyright chapters of the GPAI Code of Practice. The AI Office explained how model-use analytics can be part of post-market monitoring (Measure 3.5) and feed into systemic-risk assessment, and discussed the ‘marginal-risk’ clauses that could push competitors towards deploying unsafe models. On copyright, the obligations of Measure 1.3 were covered, including transparency towards rights holders about web crawlers and robots.txt functions.
Commission
🔥 HIGH
OpenAI and Anthropic among firms under new EU AI Act enforcement powers
Since 2 August, the European Commission has new powers to investigate, fine and impose changes on companies developing AI models. It can have AI models inspected, restrict market access and fine providers up to €15 million or 3% of worldwide annual turnover. CNBC reports that OpenAI and Anthropic are among those now falling under the new European supervisory authority as part of the AI Act rollout.
investments
🔥 HIGH
AI Gigafactories call draws 77 proposals from 16 member states
The informal call for interest for the AI Gigafactories has produced a robust response: 77 proposals to set up AI Gigafactories across 16 member states and 60 different locations. On 30 July, the EuroHPC Joint Undertaking published the official call for up to seven AI Gigafactories, backed by up to €10 billion in EU and national funding and aiming for at least €20 billion in private investment. This remains the most important sovereignty dossier for Europe’s computing power.
member states
📌 MEDIUM
EuroHPC JU launches the AI Gigafactories call
Following the adoption of Council Regulation (EU) 2026/150, the mandate of the EuroHPC Joint Undertaking has been extended to the rollout of AI Gigafactories across Europe and the promotion of quantum technologies. On 30 July, EuroHPC published the official call through which consortia will be selected to build and operate the large AI Gigafactories for training and running AI models. The call is a concrete pillar of Europe’s strategic autonomy in computing power.
technology
📌 MEDIUM
AI Office gets ready to ‘police’ frontier AI — first steps are requests for information and dialogue
The EU AI Office is gaining powers to enforce the AI Act rules on powerful (frontier) AI models. According to The Parliament Magazine, the first enforcement actions will likely not be spectacular fines, but will begin with requests for information, structured dialogue with model providers and independent evaluations. This points to a gradual, dialogue-driven approach in the first phase of AI Act enforcement, with the heaviest tools kept in reserve.
investments
📌 MEDIUM
Scaleup Europe Fund starts making investments
The Scaleup Europe Fund, which aims to help European scale-ups grow faster and compete globally with an initial target size of €5 billion, is starting to make investments. The fund is a core instrument in the strategy to financially support European ‘champions’ — including in AI and deep tech — and thus become less dependent on foreign investors. It is a concrete financing pillar of Europe’s drive for technological sovereignty.
Commission
📌 MEDIUM
Commission publishes guidelines on transparency obligations for AI providers and deployers
The Commission has published guidelines to help providers and deployers of AI systems meet the AI Act transparency obligations that applied from 2 August. The guidelines clarify who must comply with the obligations for interactive AI systems and for marking and labelling AI-generated content. Providers must design AI systems so users are informed when they interact directly with AI and add machine-readable markings; deployers must inform people when exposed to deepfakes and AI-generated content on public-interest matters.
regulation
📌 MEDIUM
‘Safer and more transparent AI’ — transparency rules enter into force
On 2 August, the new rules on transparency of AI systems entered into force, aimed at fostering trust and integrity in the information ecosystem. The Commission stresses that AI is evolving rapidly and that it is increasingly difficult to distinguish AI-generated and manipulated content from human-made, authentic content. These rules form the legal basis for labelling deepfakes, reporting chatbot use and machine-readable markings of AI content.
📈 Key Trends
- • The AI Act is now genuinely enforced — expect concrete actions from the AI Office and national authorities in the coming weeks, starting with requests for information and dialogue with model providers.
- • The Code of Practice on AI transparency won massive support (~190 signatories, from OpenAI to Lufthansa) — it is now the standard route to compliance with the marking and labelling rules.
- • Enforcement is gaining teeth vis-à-vis frontier AI: the Commission can inspect models, restrict market access and fine up to €15M/3% of turnover — OpenAI and Anthropic now fall under this.
- • The AI Gigafactories call (up to seven factories, 77 proposals from 16 member states) remains the most important sovereignty dossier for Europe’s computing power.
- • In September, the two taskforces of the Code of Practice will be launched and the Cyber Resilience Act reporting obligations (11 September) enter into force; the Scaleup Europe Fund starts investing shortly after the summer.
Wat vond je van dit bericht?