
πͺπΊ EU AI & Sovereignty Top 10
2 October 2026
This week’s EU AI & Sovereignty update β ten stories on European digital autonomy, AI policy and cybersecurity where they meet.
Sovereignty
π₯ HIGH
Dutch cabinet refuses to tell parliament about cloud-exit plans of central government bodies
The Dutch cabinet will not share risk analyses and cloud-exit plans of central government organisations with the House of Representatives. It sits right in the middle of the European debate on cloud sovereignty β and shows how uncomfortable governments still are with their dependency on foreign hyperscalers.
Cybersecurity
π₯ HIGH
Police find 110 terabytes of stolen data on ransomware group KillSec servers
An international police operation against KillSec uncovered 110 TB of stolen data. In Spain a 16-year-old suspect was arrested, and servers plus a leak site were seized. EU cybersecurity enforcement at its sharpest β and a reminder that ransomware remains a cross-border problem that no single member state can tackle alone.
Defence
π₯ HIGH
European Parliament demands a tougher EU response to hybrid threats
In a new report, MEPs warn of growing hybrid attacks on the EU, naming Russia and proxies such as Belarus as the biggest threat; China, Iran and North Korea are active too. Parliament wants more intelligence sharing β with SIAC as a central EU hub β a new cross-cutting action plan and closer NATO cooperation.
Policy
π₯ HIGH
Von der Leyen (SOTEU 2026): EU must scale compute massively and set up a European Security Council
In the annual State of the European Union debate, von der Leyen called AI and climate the “tipping points” of our time. She wants to expand European computing capacity sharply, use public-private financing for promising European startups, and announced game-changing initiatives for industrial AI (health, transport, agri-food, manufacturing, defence/space).
Innovation
π MEDIUM
Commission: Europe must scale up research and innovation to stay competitive
The EC published the 2026 edition of the Science, Research and Innovation Performance of the EU report. The core message: without extra scale and investment, the EU loses its competitiveness. This is the policy foundation under the European AI and innovation agenda β a sobering scorecard rather than a victory lap.
Cybersecurity
π MEDIUM
Dutch intelligence service AIVD warns of espionage through smart vehicles
Modern vehicles collect ever more data and are permanently online, making them an attractive target for state espionage. A digital-sovereignty risk that touches the European debate on safe connected-vehicle infrastructure β where the car is becoming a data platform on wheels.
AI Security
π MEDIUM
Belgian CCB: no AI-agent attacks observed in Belgium yet
Belgium’s Centre for Cybersecurity reports it has not yet observed attacks by autonomous AI agents. A useful baseline reading as agentic AI emerges worldwide as an attack vector β “not yet” is not the same as “not coming”.
Sovereignty
π MEDIUM
Minister: digital autonomy is not a statutory task for pension funds
Minister Vijlbrief (Social Affairs) states that digital autonomy is not part of the statutory task of pension funds. A sign that digital sovereignty in the Netherlands is still fragmented and not legally anchored β relevant to the wider EU autonomy agenda.
Geopolitics
π MEDIUM
Trump’s AI rebrand (“SI”) triggers a .si domain run in Slovenia
Since Trump wants to rebrand AI as “super intelligence” (SI, the initials of Slovenia’s top-level domain), register.si signed up 44,000 domains in September β up from under 2,000 in August (+2,100%). The Slovenian registry calls the scale “unprecedented” and is watching for cybersquatting. An unexpected geopolitical side effect of AI branding.
AI Security
π MEDIUM
OpenAI stops “adversarial distillation” campaign, but the trick kept working on Microsoft Azure
OpenAI says it dismantled a coordinated campaign copying the hidden reasoning of its models, linking a core group to people around Moonshot AI. Researchers showed the same trick kept working for weeks on cloud platforms such as Azure β pulling AI security and EU cloud infrastructure into the same story.
π Key Trends
- β’ Digital sovereignty is moving from rhetoric to procurement: cloud-exit, compute scaling and critical communication systems all land in the same policy week.
- β’ Cybersecurity enforcement is going cross-border β 110 TB seized and hybrid threats treated as structural, not incidental.
- β’ AI security becomes cloud security: stolen reasoning and agentic attack vectors both run through the same hyperscaler infrastructure Europe depends on.
- β’ The EU’s binding-rule approach keeps widening (AI Act, ProtectEU) even as national implementation β pensions, cloud transparency β stays fragmented.
Wat vond je van dit bericht?