AI Sovereignty·March 12, 2026
Why English-only benchmarks are insufficient for European AI
English-first evaluation obscures failure modes that emerge under legal, administrative, and multilingual European contexts.
Request article →Karokan
Insights
Analysis and publications on AI sovereignty, model evaluation, the AI Act, and the infrastructure layer European AI still lacks.
AI Act·March 3, 2026
EU AI Act: what organizations must prepare before August 2026
Risk classification, documentation, transparency, and human oversight must become operational disciplines well before enforcement intensifies.
Request article →Model Evaluation·February 25, 2026
Evaluating Mistral Large 3 on European legal reasoning
A structured benchmark review across risk classification, institutional document analysis, and regulatory interpretation.
Request article →AI Sovereignty·February 10, 2026
Scale AI, Turing, Mercor: the state of the human layer
The United States has scaled the human infrastructure behind frontier AI. Europe still lacks a sovereign equivalent.
Request article →Research·January 18, 2026
Multilingual RLHF: methodology and initial results
Initial findings from preference data collection across multiple EU languages, with lessons for evaluator consistency and quality control.
Request article →Research·January 6, 2026
Toward the first European benchmark for AI Act compliance
A benchmark design proposal for testing traceability, transparency, and human oversight requirements at model and system level.
Request article →