The European Commission’s AI Office has set up a consortium led by Capgemini to assess cyber risks in advanced AI systems. The work could help turn broad AI safety obligations into technical tests, though mandatory audits remain a possible future outcome, not a change announced today.
Watch Desk analysis
What happened
ITPro reports that the consortium includes Capgemini, consultancy Wavestone and France’s national metrology and testing laboratory, LNE. Its stated aims include standardising adversarial testing, vulnerability discovery and assessments of models’ offensive cyber capabilities.
Capgemini is expected to model emerging cyber threats, develop evaluation tools and conduct technical assessments on demand. ITPro says the consortium will support the AI Office’s work on systemic risks and enforcement of the EU AI Act’s obligations for general-purpose AI providers.
Why it matters
The AI Act sets out obligations for certain high-impact models, but the practical question is how their risks will be measured. A shared approach to testing could give regulators and developers a clearer basis for identifying weaknesses, including risks around automated vulnerability discovery and exploit generation.
ITPro says the initiative could lead to mandatory technical audits before high-capability models are deployed in the EU. That is a possible next step, not a rule the article says is already in force. The distinction is rather important when regulation is involved.
Our read
This is a meaningful move from describing AI risks to building the tools for assessing them. The consortium’s value will depend on whether its tests are rigorous, transparent and useful beyond a reassuring badge on a compliance checklist. We’ll be looking for the methods and evidence, not just the launch announcement.
What to watch
- Which testing methods and evaluation tools the consortium develops.
- Whether the AI Office publishes how assessments will be used in practice.
- Whether technical audits become mandatory, and which models or providers would be covered.
Discussion spark: Should the EU require independent technical audits of high-capability AI before deployment, or should developers be allowed to rely on their own testing?
Sources and evidence
- Source update (10 November 2026, 08:56 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.