Superintelligence Europe — No. 007

The European Parliament passed its plenary position on the Digital Omnibus amendment to the AI Act by 569 votes in favour, 45 against, and 23 abstentions — an unusually large margin for a piece of legislation touching both civil liberties and business competitiveness. The position now opens trilogue negotiations with the Council of the EU, with a final compromise expected as early as April.

The headline provision: a new outright prohibition on AI “nudifier” systems — tools that create or manipulate images to depict identifiable real people in sexually explicit or intimate ways without their consent. The ban was not in the Commission’s original proposal; it was pushed by Parliament, led by co-rapporteur Michael McNamara (Renew, Ireland). The vote came the same day that the Amsterdam court ruled against Grok — a convergence that made the political symbolism impossible to miss. Beyond the ban, Parliament also voted to push the application deadline for most high-risk AI systems to December 2, 2027, giving companies and national authorities more time to prepare for obligations that were already running behind schedule. SME flexibility provisions and data bias correction safeguards were also included.

The Amsterdam District Court issued a preliminary injunction on Thursday prohibiting xAI and its Grok chatbot from generating or distributing sexual imagery of persons partially or wholly stripped naked without their explicit consent anywhere in the Netherlands. Non-compliance carries fines of €100,000 per day per defendant. The court also ordered that Grok cannot be offered on the X platform while in breach of the ruling. The case was brought by Offlimits, a Dutch nonprofit focused on combating online sexual abuse, working with the Victims Support Fund.

xAI had argued at the March 12 hearing that it had implemented stringent safeguards since January 20, 2026 and that guaranteeing zero misuse was technically impossible. The court found that argument directly undermined by Offlimits’ evidence: on March 9 — the same day xAI sent its categorical denial to Offlimits — Offlimits was still able to generate a sexualised video of an existing person from a single uploaded photograph using Grok, without any consent verification. The court ruled that as an internet intermediary with control over Grok’s functionalities, xAI bears responsibility for preventing the generation of unlawful images — regardless of whether individual users issue the prompts.

Verne, Pony.ai, and Uber announced on Thursday a three-way partnership to launch Europe’s first commercial robotaxi service, starting in Zagreb, Croatia. Public road testing using Pony.ai’s Gen-7 autonomous driving system, deployed on the Arcfox Alpha T5 robotaxi platform, is already underway. Fare-charging services are in preparation. Uber will invest directly in Verne as a strategic partner and integrate the service into its global ride-hailing network, alongside Verne’s own app.

Verne is a Croatian startup backed by Rimac Group — the company whose founder Mate Rimac has spent seven years working toward exactly this moment. Pony.ai brings proven commercial-scale autonomous driving technology from China, where it has already reached unit economics breakeven in Guangzhou and Shenzhen. The partnership structure is deliberate: Pony.ai provides the technology, Verne owns and operates the fleet, and Uber provides distribution at scale. Verne will also be responsible for obtaining European regulatory approval and coordinating cross-market expansion — with plans to grow to thousands of robotaxis over the next few years.

Politico published a detailed infrastructure analysis on Thursday documenting the growing gap between the EU’s stated ambition — thousands of new AI data centres under the AI Continent Action Plan and InvestAI programme — and the physical reality of where those facilities can actually be built and powered. Grid capacity constraints, high European energy costs, and increasingly vocal local resistance are combining to create a practical ceiling on how quickly the EU can build the sovereign compute infrastructure its AI strategy depends on.

The piece builds on the broader picture established by Wired’s grid investigation earlier this week — but adds the dimension of public opposition, which is emerging as a factor in planning decisions across multiple member states. Data centres are large, energy-intensive, water-cooled facilities. In communities where energy costs are already a political pressure point, their arrival is not always welcome. The collision between Brussels’ sovereignty rhetoric and local planning realities is the infrastructure story beneath the infrastructure story.

Credible accounts and publications that drove Thursday’s European AI conversation. Filtered for genuine signal.