No. 89
2026-10-16
the AI pulse — brewed, not raw
 AI CRITIQUE
Gri tonlu bir illüstrasyon, bir Almanya haritası üzerinde stilize edilmiş bir tuğla bina ve bir masada dizüstü bilgisayar görüntüsünü betimlemektedir.

Seven German states share LLMoin AI assistant

Read the article    English (automatic translation)

Summary

Dataport, a public IT provider jointly owned by six German states, reported that its LLMoin AI assistant has reached 100,000 users across seven states and 17 municipalities. The company positions the assistant as a reusable, multi-tenant tool for public administration.

Why it matters

This represents one of the clearest real-world examples of cross-border shared AI resources in the European public sector. It demonstrates that a single product can serve different administrative levels while maintaining data isolation for each institution. Furthermore, it highlights a strategic move toward sovereign, model-agnostic public AI adoption in the EU, serving as a deliberate countermeasure against uncontrolled 'shadow AI' usage among civil servants.

Highlights

  • 100,000 users across 7 states and 17 municipalities; this figure was approximately 60,000 just ten months ago
  • Multi-tenant architecture with Active Directory integration and institution-based data isolation for each state
  • Model-agnostic design with servers hosted in the EU; the roadmap lists a GPT-5.5 upgrade for version v1.23

Critical take

Dataport's headline figure of '100,000 users' is surprisingly vague. The company does not clarify whether these are registered accounts or active users, making it difficult to verify or benchmark the adoption claim. The side-by-side comparison with France's DINUM 'L'Assistant' tool in the article is also weak. The two tools serve different administrative scopes, and the article itself acknowledges that their usage volumes are not directly comparable.

Bu kritik, yerel yapay zeka modeli Qwen3.8-27B tarafından yazılmıştır. Haberi ben seçtim; kritik metnini yayımlanmadan önce tek tek denetlemiyorum.This critique was written by the local AI model Qwen3.8-27B. I chose the story; I do not review each critique before it is published.
Back to the issue