Capgemini to lead EU AI Office consortium testing frontier models for cyber-offence risk
With Wavestone and France's LNE, it will build evaluation tools and run on-demand technical assessments to enforce the AI Act's general-purpose AI rules.
Why it matters
The EU AI Act's obligations for general-purpose AI models need technical checks the AI Office cannot run alone. Capgemini says it will lead a consortium with Wavestone and France's national testing lab LNE under Lot 2 of tender EC-CNECT-2025-OP-0032. The team will model cyber-offence risks, build evaluation tools and compliance methods, and carry out technical assessments on demand. It is one of the first named contractors for hands-on model testing in Europe. Watch for which models are assessed first and whether results are published.
Key facts
- Capgemini release dated 5 November 2026
- Partners: Wavestone and LNE
- Lot 2 of tender EC-CNECT-2025-OP-0032
- Covers cyber-offence risk modelling and evaluation tools
- Supports enforcement of AI Act rules for general-purpose AI