The EU's AI governance apparatus shifted in two directions simultaneously this week: legislators agreed to soften compliance burdens on smaller operators, while the Commission signalled it may soon push sensitive government workloads off US cloud infrastructure entirely. Both moves arrive as the May 27 Tech Sovereignty Package approaches, a bundle of legislation that will define the architecture of Europe's digital autonomy for the rest of the decade.
Policy & Regulation
AI Act simplified for SMEs and general-purpose models. On May 7, the European Council and Parliament reached a provisional agreement to streamline several AI Act obligations. The revisions extend the deadline for national AI regulatory sandboxes to August 2027 and compress the grace period for labelling AI-generated content from six months to three, tightening the synthetic content rule while loosening the infrastructure burden. The deal reflects sustained industry pressure to reduce compliance costs ahead of the August 2 full applicability date. (European Council)
Commission opens consultation on AI transparency guidelines. On May 8 — one day after the Council deal — the European AI Office launched a stakeholder consultation on draft guidelines for AI transparency obligations. From August 2026, EU users must be informed when interacting with AI systems or encountering synthetic content. The draft guidelines set out what disclosure looks like in practice; the accompanying Code of Practice, drafted by independent experts, is expected in June. Comments are open until June 3. (European Commission)
EU weighs restricting US cloud for sensitive government data. The Commission is considering rules that would bar EU public administrations from processing sensitive government data on US hyperscaler platforms, sources told CNBC on May 7. The proposal is likely to appear in the Cloud and AI Development Act (CADA), part of the Tech Sovereignty Package due May 27. Amazon, Microsoft, and Google currently hold more than 65% of EU cloud market share. CADA is also expected to introduce a legal framework for EU-based high-performance compute capacity tied to strategic autonomy goals. (CNBC)
DMA passes first review — scope stays, enforcement intensifies. The Commission published its mandatory first review of the Digital Markets Act on May 3, finding the regulation "fit for purpose" after two years in force. The review declines to add generative AI or social networking to the list of designated core platform services for now, calling it premature. The focus going forward will be enforcing existing obligations more aggressively, particularly on cloud designation, where US hyperscalers have so far avoided gatekeeper status despite dominating the market. Meanwhile, Apple and Meta were fined €500 million and €200 million respectively in April for non-compliance. (TechPolicy.Press)
DSA second wave extends obligations broadly. The Digital Services Act's second enforcement phase began in Q2 2026, expanding obligations that previously applied only to Very Large Online Platforms to a much wider set of companies. The EU simultaneously signed a DSA cooperation arrangement with Japan's Ministry of Internal Affairs and Communications — part of a pattern of building regulatory coalitions with allied democracies. (European Business Magazine)
Capital & Investment
EIC deploys €146.5M under STEP Scale Up. Eight European startups will receive equity investments of between €10M and €30M each under the European Innovation Council's STEP Scale Up scheme, targeting scale-up rounds of €50M–€150M. The cohort spans deep tech and industrial sectors, with private co-investment expected to multiply the public capital. The EIC has positioned this instrument as a bridge for companies that have outgrown early-stage grants but struggle to attract late-stage European institutional capital. (EIC)
Legora extends Series D past €500M total. Swedish legal AI startup Legora raised a €42M extension to its Series D, pushing total funding above €500M. The raise signals continued investor appetite for vertical AI in professional services — a segment where European founders have built defensible positions against US incumbents. Legal AI is increasingly viewed as one of the few areas where a European-headquartered company can reach category leadership before a Silicon Valley player does. (EU-Startups)
UK fintech Ebury raising £550M with Santander increasing stake. London-based cross-border payments company Ebury is in the process of raising approximately £550M, with its existing backer Santander set to increase its stake to 55%. The round would value Ebury at a significant premium to its last public valuation and marks one of the largest European fintech raises of the year. It also continues a pattern of incumbent banks deepening positions in payment infrastructure they initially acquired minority stakes in. (EU-Startups)
Pre-seed ecosystem remains active. Warsaw-based Montis VC reached first close on a €50M fund targeting energy and industrial transition startups — one of several new European pre-seed vehicles launched in Q1–Q2 2026. DFF Ventures also closed Fund III at €70M, signalling that LP appetite for early-stage European tech remains strong even as Series B+ rounds face more scrutiny. The divergence between buoyant pre-seed and compressed growth stages is a structural feature of the current market, not a blip. (EU-Startups)
Talent & Workforce
Europe emerging as third pole for AI talent. A Euronews analysis published April 29 found that visa restrictions in the US and narrowing pipelines from China are redirecting significant AI talent flows toward Europe. The UK, Germany, France, and the Netherlands are the primary destinations. The structural advantage is real: relative openness, proximity to strong research universities, growing clusters of well-funded AI labs. But European governments have done little to actively capitalise on the shift. Talent acquisition remains largely market-driven rather than strategically coordinated. (Euronews)
One to Watch
Tech Sovereignty Package — May 27. The Commission is expected to table its Tech Sovereignty Package on May 27, bundling the Cloud and AI Development Act (CADA) with Chips Act 2.0 and possibly other measures. CADA's scope will determine whether Europe is serious about digital infrastructure independence or building another framework that stops short of structural change. The key question: will it include binding restrictions on US hyperscalers for public-sector workloads. The political temperature around this is higher than most Commission packages; US trade pressure, AI competitiveness anxiety, and genuine sovereignty concerns are all pulling in the same direction for once. Watch the CADA text closely when it drops. (European Parliament Legislative Train)