European Union Pressed AI Firms on Safety Standards
The European Commission has requested safety data from over 30 firms as it enforces new foundational model regulations.
Updated on Oct. 11, 2026 in Artificial Intelligence
Live Poll
Do you believe governments should strictly regulate the safety and security of advanced artificial intelligence models?
European Commission official Henna Virkkunen confirmed that the EU AI Act now enforces safety and security mandates for foundation and frontier AI systems. In late August 2026, the commission issued information requests to more than 30 AI companies to evaluate compliance with these newly established measures.
Why it matters
The regulation aims to manage emerging risks associated with highly capable AI models by mandating rigorous safety documentation and oversight. This action marks the commission's shift toward active enforcement of the tiered risk-based framework established by the AI Act.
The AI Act classifies systems into four distinct risk levels to determine regulatory requirements. These mandates require companies to conduct formal risk assessments, solicit input from external experts, and maintain continuous monitoring of their foundation models.
The players
Henna Virkkunen
A European Commission official responsible for affirming the safety and security oversight mandates of the EU AI Act.
European Commission
The executive branch of the European Union responsible for drafting and enforcing legislation, including the comprehensive AI Act regulatory framework.
The details
The regulatory process requires developers to submit technical documentation proving they have identified potential safety hazards. The AI Act mandates the integration of risk mitigation strategies, including the use of external subject-matter experts to audit system outputs and the implementation of ongoing monitoring protocols to track model performance after deployment. This tiered approach is designed to prevent the market release of systems that lack verified safety controls.
Timeline
December 2023: The AI Act reached provisional status.
February 2024: Four risk levels for foundation and frontier AI systems were established.
May 2024: The European Commission formally approved the AI Act.
Late August 2026: The commission sent information requests to over 30 AI companies.
The Tech Race
This enforcement action follows the framework established by the EU AI Act, which separates model oversight into four specific risk tiers. It aligns with broader international efforts, including discussions at the United Nations, to standardize safety protocols for frontier AI models.
The regulatory shift forces AI developers to prioritize safety documentation and external auditing before releasing new foundational systems in the European Union. Users should expect a more standardized, albeit potentially slower, pace of AI deployment as firms align their product roadmaps with these mandated risk-assessment tiers.
The takeaway
The move signals a transition from the legislative design of the AI Act to active industry compliance monitoring. Observers should track the responses of the 30 notified companies to see which firms modify their deployment timelines to meet the new four-tier risk mandates.
Further reading
Explore the broader implications of global regulatory trends in our Artificial Intelligence section.
Live Poll
Do you believe governments should strictly regulate the safety and security of advanced artificial intelligence models?
