Superintelligence Europe — No. 014

Euronews published its most detailed analysis yet of the Programme for Agile and Rapid Defence Innovation — AGILE — on Tuesday, bringing the European Commission’s March proposal into the widest English-language circulation since its announcement. The programme commits €115 million as a pilot to fast-track disruptive defence technologies from the lab to operational armed forces use. Its primary focus areas are AI systems for military decision-making and situational awareness, autonomous systems, drone technologies, quantum computing, and advanced robotics. Unlike the EU’s existing defence innovation frameworks — the European Defence Fund, EUDIS, EDIP — AGILE is explicitly designed around speed rather than process. Grants will cover up to 100 percent of eligible costs. Funding decisions are expected within four months of application. Technologies are expected to reach armed forces within one to three years of award.

The programme is designed specifically for the “new defence players” — start-ups, scale-ups, and SMEs — that move at commercial technology speed rather than traditional defence procurement pace. Between 20 and 30 projects are expected to receive individual grants of €1 million to €5 million. Companies from EU member states, as well as Norway and Ukraine, will be eligible to apply when the first call opens early 2027. EU Defence Commissioner Andrius Kubilius framed the programme explicitly: Europe can no longer rely on defence innovation cycles measured in years when modern warfare now depends on technologies that need to be developed, tested, and deployed in weeks or months. Ukraine’s war has made that reality visible. AGILE is the institutional response.

The Commission described the programme as designed for “the ‘New Defence’ players, the startups and tech innovators who move at high speed” — a deliberate signal that AGILE is not intended for the large traditional defence primes that have historically dominated EU funding. The administrative burden has been reduced: single companies can apply without forming multinational consortia, retroactive funding is permitted, and grants cover up to 100 percent of eligible costs.

The European AI Office opened a targeted consultation on Tuesday seeking input on how to measure and mitigate the energy footprint of AI models and data centre systems. The consultation window runs from April 7 to May 15, 2026. This is the first formal EU regulatory step toward establishing a measurement methodology for AI energy consumption — a question that has been present in policy discussions since the emergence of large language model training but has not previously been the subject of a targeted Commission consultation.

The timing connects directly to Monday’s Futura-Sciences analysis of France’s ADEME data: 352 data centres consuming 2.2 percent of France’s national electricity, with a trajectory to 500 by 2030. The ADEME report made visible the environmental cost of European AI infrastructure at the national level. The EU consultation is the regulatory mechanism for addressing that cost at the continental level. The two are not coordinated — ADEME is a French national agency, the EU AI Office is a Commission body — but they arrived on consecutive days and point at the same underlying tension: European AI’s sovereignty goals and its environmental commitments are increasingly in conflict, and someone needs to quantify the trade-off before it can be managed.

The UK Information Commissioner’s Office published “Recruitment Rewired” — the most comprehensive regulatory audit of AI in UK hiring to date — and it drew its widest attention on Tuesday as Staffing Industry Analysts and UK tech media amplified the findings. The ICO engaged with more than 30 major employers between March 2025 and January 2026 to understand how automated decision-making is being used in recruitment. Its key finding: many employers are relying on solely automated decisions — systems operating without meaningful human involvement — to make or significantly influence hiring outcomes. That places those decisions within the scope of UK GDPR’s sole automated decision-making provisions, requiring a range of safeguards that the ICO found are frequently absent.

The range of AI tools identified in use is wider than most public discussion has acknowledged. Beyond CV screening and ranking, employers are using AI-powered behavioural games and psychometric assessments, sentiment and emotional analysis of interview transcripts, personality prediction from tone and language patterns, and scoring of written responses to interview questions. The ICO wrote to 16 organisations identified as likely to be running solely automated decisions affecting candidates, and all 16 have now committed to acting on the recommendations. Separately, the ICO launched a consultation on updated draft guidance on automated decision-making and profiling, open until 29 May 2026 — running concurrently with the EU AI Office’s energy consultation.

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