The bloc introduces evaluation capacity, joint procurement options, and sovereignty measures amid disruptions to advanced models and shifting policy signals.
Following a period of instability triggered earlier when the United States imposed limits on foreign access to some of the country’s most advanced AI models, the European Commission has introduced a new strategy addressing both the risks and strategic opportunities tied to advanced AI in cybersecurity.
The bloc’s move on 7 Jul 2026 signals the need for a sharper focus on resilience after recent disruptions in global AI access. Central to the plan is a set of contingency measures designed to safeguard European stakeholders in scenarios where access to leading AI systems is restricted.
Tech Commissioner Henna Virkkunen has announced plans to establish an “EU evaluation capacity” for advanced AI systems, expected to be operational by 2027. This initiative is intended to provide AI developers with a European-based pathway for conducting the independent assessments required under the EU AI Act, while also strengthening enforcement capabilities within the bloc’s AI Office.
Alongside this, the Commission has outlined a “blueprint” to guide how AI providers can grant secure and compliant access to European entities. The proposal also explicitly acknowledges the risk of access being withdrawn, noting that EU institutions and member states could intervene collectively, including through joint procurement of model access if necessary.
The broader policy direction reflects growing concern in Europe over-reliance on US-based AI providers.
As recent as in late June, Reuters had reported that European firm have already begun diversifying their AI partnerships to reduce exposure to sudden policy shifts. The Commission’s latest plan builds on its wider Tech Sovereignty Package, which includes initiatives such as AI Factories and proposed Gigafactories aimed at strengthening domestic infrastructure.
Analysts at Chatham House have warned that repeated reversals in US policy send inconsistent signals about AI governance, raising doubts about the reliability of continued access to frontier models. The latest reactions in the EU and worldwide suggest an effort to address those concerns directly by reinforcing strategic autonomy in the AI landscape.
