Superintelligence Europe — No. 009

The European Commission has now missed its second consecutive deadline for publishing critical guidance that businesses need to comply with the AI Act’s high-risk system requirements. The February 2, 2026 deadline — set under the Act itself — came and went without the practical implementation guidelines that providers and deployers of high-risk AI systems had been counting on to structure their compliance programmes. This follows the Commission’s failure to meet the same category of obligations in the prior cycle.

The timing creates a structural contradiction at the heart of EU AI governance. On one hand, the Parliament voted this week to delay high-risk AI obligations to December 2027 — precisely because standards and guidance were not ready. On the other, thousands of businesses are currently attempting to prepare compliance programmes in the absence of the official interpretation that would tell them what compliance actually looks like. The Digital Omnibus delay buys time at the legislative level. The missing guidance creates uncertainty at the operational level. For companies with compliance cycles of twelve months or more, the gap between those two things is already costing money.

The week’s most technically significant European AI funding story was picked up in this Sunday’s TNW weekly recap: Kandou AI, a fabless semiconductor company headquartered in Saint-Sulpice, Switzerland and spun out of EPFL in 2011, closed an oversubscribed $225 million Series A round led by Maverick Silicon, with strategic participation from SoftBank, Synopsys, Cadence Design Systems, and Alchip Technologies. The round values the company at $400 million.

The company’s proposition sits at the most contested technical frontier in AI infrastructure: the interconnect bottleneck. As AI clusters scale to tens of thousands of GPUs, the constraint on performance is no longer raw compute but the speed and efficiency with which data moves between processors and memory. The prevailing industry response has been to shift from copper to optical interconnects — a transition that is expensive, complex, and still maturing. Kandou AI’s bet is that copper, if signalled correctly using its patented Chord technology, can match or exceed the performance of optics at a fraction of the cost and power consumption. Its claimed figures — ten times lower power than optical alternatives, bandwidth beyond 448 gigabits per second — if they hold at commercial scale, would represent a significant challenge to the billions currently being invested in optical alternatives.

The European Financial Review published a detailed risk analysis on Sunday documenting how AI is reshaping the cyber threat environment for European banks, insurers, and financial market infrastructure heading into the second half of 2026. The piece covers three distinct attack vectors that AI has materially worsened: deepfake-enabled fraud targeting executive authorisation processes, automated phishing campaigns using personalised financial data at scale, and model poisoning attacks targeting the AI systems that financial institutions are themselves deploying for fraud detection and credit scoring.

The analysis lands against a specific regulatory backdrop. DORA — the Digital Operational Resilience Act — entered full application for EU financial entities in January 2026, requiring robust ICT risk management, incident reporting, and third-party oversight. But DORA was designed before AI-enabled attacks reached their current sophistication. The report argues that governance frameworks need updating faster than the regulatory cycle can accommodate, and that the AI Office’s focus on the AI Act’s high-risk classification framework is not addressing the financial sector’s most immediate AI risk exposure.

Credible accounts and publications that shaped Sunday’s European AI conversation. Filtered for genuine signal.

This was not a typical week. Seven days produced the sharpest legislative action the European Parliament has taken on AI harm, the first binding judicial order against an AI image generator on the continent, the announcement of Europe’s first commercial robotaxi service, and the most explicit transatlantic confrontation over AI regulation of the year so far. For readers coming to this briefing fresh, here is the shape of the week: