🤖 EU AI & Sovereignty Top 10
29 October 2026
Theme of the day: Europe does not have a money problem, it has a middle-of-the-curve problem. Four of today’s ten items are about capital — an equity facility the Commission is only exploring, €21 billion in SAFE loans nobody has claimed, a mid-segment funding gap that grew while early-stage rounds boomed, and a Norwegian LP fund filling part of the hole. The other six are the things capital cannot buy: rules for algorithmic management that the AI Act does not cover, and liability for a deepfake that a national court now has to sort out.
capital
🔥 HIGH
Commission is still in an “exploration phase” for a European tech equity fund
The Commission is sounding out a new investment instrument meant to finance strategic innovation in the EU: an actual equity pot for scale-ups rather than another loan facility. The article’s tagging ties it to the European Competitiveness Fund, the European Investment Bank, the Scaleup Europe Fund, own resources and the portfolio of tech-sovereignty commissioner Henna Virkkunen. That is a lot of institutional furniture for something that is not yet a proposal. The honest reading: Brussels has recognised that Europe grows companies to Series A and then loses them, but has not decided whether it wants to be an investor or a grant-giver. The instrument only matters if it can mobilise billions, and nothing in the exploration phase suggests a number.
strategy
📌 MEDIUM
“The Power Dividend”: Europe has more leverage than market size alone
A Rapporteur opinion argues that the EU keeps reaching for regulation where it should reach for leverage. China’s advantage is not that it is a bigger market, it is that Beijing treats market access as a bargaining chip — and Europe has the largest single market in the world and barely uses it that way. The piece lands in the middle of a live argument about how the EU should answer Chinese export licensing and American platform power. It is also the intellectual frame for item #1: if you want to bargain, you need something to bargain with, and a funding instrument is one of the few things the EU still controls outright.
financing
🔥 HIGH
€21 billion in SAFE loans is still unclaimed — and Brussels made the terms stricter
After clearing the Hungarian plan, €21 billion of the SAFE loan capacity remains unused, and the Commission is now hunting for takers. The second round is harder to sell precisely because Brussels tightened procurement rules on the back of the first — conditionality that was meant to protect the money now raises the cost of taking it. This is the unglamorous half of European security and tech sovereignty: the capacity exists, the demand does not, and a loan that nobody draws down does not build a chip plant. Watch whether member states convert SAFE into grants or defence-only spending; that would quietly end its use as a tech instrument.
labour
🔥 HIGH
MEP Li Andersson: an Amazon warehouse visit shows why AI needs workplace rules
After visiting an Amazon site, MEP Li Andersson calls algorithmic management “the core challenge” — and points out that the AI Act does not cover it. Her argument: the question is not whether an algorithm assigns work, rates workers or fires them, but whether anyone negotiated the terms. That places the dossier in employment and social security law rather than AI regulation, which is exactly why it has been slow: the AI Act was written around products and risk classes, not around the employment relationship. For anyone building HR analytics in Europe this is the item to watch — the compliance question will arrive through labour law and works councils, not through a conformity assessment.
payments
📌 MEDIUM
Digital euro: member states push for a final deal before the end of the year
Member states are aiming to close a definitive agreement on introducing the digital euro before the end of 2026, which would move the project from design into its final phase. This is the most explicit sovereignty dossier in payment infrastructure: every card transaction in Europe currently clears through two American networks, and the digital euro is the only serious attempt to build a European alternative at the settlement layer. The remaining fights are the familiar ones — holding limits, privacy of small transactions, and how much of the ledger the ECB itself gets to see. The deadline matters: a deal this year puts the legislative framework in place before the next Commission cycle.
privacy
🔥 HIGH
noyb files complaints against Scale AI with the Dutch DPA and three sister regulators
Privacy group noyb has filed a complaint against AI data company Scale AI with the Autoriteit Persoonsgegevens, alongside sister regulators in three other member states, over how the company monitors the contractors who label and evaluate data for the big model providers. Reporting by Rapporteur adds that the complaints centre on workplace surveillance: researchers argue the platform collects worker data without a valid legal basis. The structural point is bigger than one company. This is the labour-intensive underlayer of the AI industry — it sits outside the AI Act and entirely inside the GDPR, which means enforcement depends on data protection authorities rather than on AI supervisors. A Dutch complaint makes the AP the test case for how that layer gets regulated.
enforcement
📌 MEDIUM
Commission accepts Temu’s action plan to remedy its DSA breach
The Commission has accepted Temu’s action plan to repair the breach it had already been fined €200 million for. The plan has five concrete parts: a platform-specific risk analysis, reliable metrics to measure whether the measures actually work, an analysis of the limits of content moderation, research into whether the recommender system amplifies illegal goods, and a study of how sellers abuse Temu’s features. Temu has four months for an interim risk assessment covering the period up to 30 November 2026 — which is next month, for a plan accepted this week. The signal is the shift in method: the DSA is becoming a supervisory regime with remediation plans and follow-up dates, not a rulebook that gets read once.
markets
🔥 HIGH
Opinion: scaling, not starting, is the real test of Europe’s AI boom
The numbers in this EU-Startups piece are the most useful thing published about European AI funding this week. Series A to C rounds grew roughly 66% in the first nine months of 2026 — and rounds between €22.2 million and €88.9 million fell by about 8%. That is a barbell: plenty of seed and Series A money, plenty of mega-rounds, and a hole in the middle where a company either becomes a European champion or gets bought. It is the structural weakness that the equity facility in item #1 is designed to address, and it explains why the Commission is talking about instruments at all. The uncomfortable implication: money at the ends of the curve does not fix a gap in the middle.
capital
📌 MEDIUM
Norway’s Investinor raises a new LP fund amid the European scale-up push
Investinor, Norway’s state-backed investor, is raising a new fund from limited partners, in line with the broader European move to steer more capital toward scale-ups. Norway is not in the EU, which is exactly why this belongs in a European list: the Nordic capital base is one of the pools that Dutch and German funds lean on, and a state investor going back to LPs is a signal about where the marginal euro for European growth companies is coming from. Compare it with item #3: public loan capacity sitting unused while a state LP fund finds private takers. The appetite is there; the instruments are not always in the right shape.
deepfake
📌 MEDIUM
Sánchez’s wife sues the Spanish opposition over an AI deepfake
Begoña Gómez is suing the Spanish opposition over an AI-generated deepfake, on grounds of identity theft. According to Rapporteur the case also touches the European tech rulebook: it is the first practical test of the AI Act’s transparency obligations in a political deepfake conflict inside a member state, rather than at the EU level where the rules are written. The legal route is deliberately not the AI Act — identity theft and defamation are older and faster instruments — but the outcome will set expectations for how deepfakes get litigated in Europe for years. Anyone shipping image or voice generation should read this one: labelling obligations are cheap compared to being the test case.
📈 Key Trends
- • The bottleneck is the middle of the funding curve. Series A–C up 66%, mid-sized rounds down 8%: Europe funds starts and mega-rounds, and loses the companies in between.
- • Capacity without demand is not sovereignty. €21 billion in SAFE loans is unclaimed while the Commission only explores an equity fund — the money exists, the takers do not.
- • The AI Act stops at the workplace door. Algorithmic management, the clearest AI power imbalance in daily life, is being handled through labour law and data protection instead.
- • Enforcement is now remediation, not rulemaking. Temu’s five-part action plan with interim deadlines shows the DSA operating as a supervisory regime with follow-up.
- • Deepfakes will be fought with old law first. Identity theft and defamation move faster than AI transparency rules, and the first national case will set the tone for the rest.
- • Payment infrastructure is the quietest sovereignty dossier. A digital euro deal this year would put a European settlement layer in place before the next Commission term.
Wat vond je van dit bericht?
