EU AI Act Enters Its Toughest Phase as Europe Weighs Its Own Rules
The EU AI Act's obligations for general-purpose AI models, due on 2 August, land in a market where European startups raised about $52bn in venture capital last year against $209bn for their US counterparts.

Two years after it entered into force on 1 August 2024, the AI Act is moving from legislation into enforcement. The next tranche covers general-purpose AI models, the systems underneath most consumer chatbots and enterprise tools. It arrives with the bloc still arguing about whether it has the balance right.
The European Commission's AI Office drafted the rules through a broad code of practice group. Eoghan O'Neill, a senior policy officer at the AI Office, told TNW Conference in Amsterdam on 20 June that members included major model providers, civic society organisations, NGOs, academics, AI safety experts, SMEs and European industrial giants.
"This is a big, sophisticated technology, and we want to get it right," O'Neill said. "We need specific obligations to capture some of the most impactful or potentially harmful models under the AI Act."
He called the drafting process "a big tent with all of those voices from the stakeholder community". The Commission planned to finalise its rules for general-purpose AI in July, he said, with the European Parliament then adopting its position on the standards. That timeline matters. The obligations were expected to take effect on 2 August, and any slippage feeds straight into the compliance planning of every company selling AI into Europe.
The AI Act sits inside a wider stack. GDPR Local, which tracks EU tech compliance deadlines, notes that the AI Act, the Digital Services Act, the Digital Markets Act, the NIS2 Directive, the Data Act and the Cyber Resilience Act each impose distinct but overlapping requirements, and that fines can reach up to 7% of global turnover. An AI system processing personal data must satisfy both the AI Act and the GDPR. If it touches critical infrastructure, NIS2 cybersecurity duties apply too. That overlap, not any single rule, is what compliance teams are actually budgeting for.
Pushback from inside the bloc
Calls to slow the roll-out have intensified from some key stakeholders, including Swedish Prime Minister Ulf Kristersson, Bosch CEO Stefan Hartung and the tech lobbying group CCIA Europe, whose members include Alphabet, Meta and Apple. The objection is not that AI should go unregulated. It is that Europe's rules are being layered on top of a market already fragmented by language, national law and capital scarcity.
Fabrizio Del Maffeo, CEO of the Netherlands-based chip company Axelera AI, put it bluntly at the same conference. "Europe is not the United States," he said. "We have many languages, many markets, and many regulations, both European and local. And these are stifling growth because they create borders, making it difficult for companies to expand." Del Maffeo said his company had signed the petition for EU Inc, a proposal for a standardised legal entity for startups operating across member states. Commission President Ursula von der Leyen described it at Davos in January as combining corporate law, insolvency, labour law and taxation in one framework.
The funding numbers behind that complaint are stark. European startups raised about $52bn, or €44bn, in venture capital last year, according to figures cited at the conference. US counterparts attracted $209bn, or €177bn. Europe accounts for just 8% of the world's scaleups, against 60% in North America, and no EU-founded startup in the past 50 years has passed a €100bn valuation.
Peter van der Putten, director of the AI Lab at software firm Pegasystems, argued the EU needs to become more attractive for both domestic and international investment. Regulation could be adjusted, he suggested, so that funding leaving the US flows into Europe instead.
Not everyone frames the gap as purely regulatory. Elise de Reus, co-founder of Cradle, pointed to European engineers returning from Big Tech roles in the US, drawn by purpose-driven work and quality of life. "We're also maybe a little bit too modest," she said. "We should measure happiness, not GDP, which is not a sustainable metric. I don't think we should copy and paste the American system."
For now, the compliance calendar runs ahead of the political argument. Companies selling general-purpose models into the EU must prepare for obligations that, on the Commission's own account, are still being finalised.
Sources
2- 01EU Tech Regulations: Compliance Deadlines and Obligations for 2026EN
- 02'Europe is not the US': Tech insiders call for smarter AI rulesEN
All figures and quotations in this text come from the sources listed below.
Content prepared by the editorial team with AI assistance.
Comments
0- No comments yet — be the first.