Summary
- The Commission plans EU-level work on AI model evaluation, structured access, and secure testing.
- Critical infrastructure cyber hygiene and open-source security are included in the plan.
- The policy connects AI security with the AI Act, Cyber Resilience Act, NIS2, and Cyber Solidarity Act.
The European Commission has set out a plan to manage the cybersecurity risks and opportunities of advanced artificial intelligence, placing model evaluation, critical infrastructure protection, open-source security, and European cyber capability inside a single policy programme.
The plan says advanced AI can improve cyber defence, while also being misused to identify vulnerabilities, automate attacks, and increase the scale and speed of incidents. The Commission’s published outline includes work on AI model evaluation, structured access to advanced AI capabilities, a secure testing platform, stronger vulnerability-fixing practices, and a campaign to secure critical open-source software.
The Commission says the AI Act requires advanced AI models to be evaluated and their risks assessed before they are placed on the EU market. It plans to support an EU evaluation capability and work with ENISA on a blueprint for structured access to advanced AI capabilities for cybersecurity. ENISA and the Joint Research Centre are also expected to create a secure platform to test AI for cybersecurity, including simulated environments for operators in critical sectors.
The plan spans more than model governance. It connects advanced AI to vulnerability management, cyber hygiene, critical infrastructure, security by design, open-source software, European AI infrastructure, and market development. The Commission also links the work to existing EU frameworks, including the AI Act, Cyber Resilience Act, Network and Information Systems Directive, and Cyber Solidarity Act.
That wider regulatory setting reflects how AI risk is likely to arrive inside organisations. Exposure will rarely be confined to a single chatbot deployment or model procurement. AI capabilities can affect secure development, vulnerability discovery, phishing, incident triage, code generation, third-party tooling, and the speed at which attackers adapt. Where regulated operators use AI in security operations or software development, assurance, logging, change management, and model access become part of operational resilience.
The Commission’s emphasis on structured access also carries practical weight. Defensive use of advanced models depends on whether public authorities, researchers, and operators can test and use high-capability systems safely, without turning access into an unmanaged risk. A European blueprint could help define the relationship between model providers, regulators, and sectors that need to evaluate capabilities before relying on them for security-sensitive tasks.
The open-source element gives the plan a supply chain dimension. Critical open-source components sit beneath public services, financial platforms, healthcare systems, industrial tools, and commercial software. AI may help identify and remediate flaws more quickly, but it can also increase the volume of findings, false positives, and rushed patching activity. Poorly governed acceleration can create operational risk even when the underlying intention is defensive.
The plan remains a policy framework rather than a completed operating model. Its impact will depend on funding, authority, sector adoption, and the extent to which practical support reaches organisations beyond large technology providers and national cyber agencies. The EU is nevertheless placing advanced AI inside cyber resilience, product security, and critical infrastructure policy, rather than leaving it as a detached innovation agenda.
