A PYMNTS Company

Pacing Frontier AI Models – The EU Regulatory Perspective

 |  September 25, 2026
Pacing Frontier AI Models – The EU Regulatory Perspective

By: Stijn Huijts (The Platform Law Blog)

    Get the Full Story

    Complete the form to unlock this article and enjoy unlimited free access to all PYMNTS content — no additional logins required.

    Subscribe to our daily newsletter, PYMNTS Today.

    By completing this form, you agree to receive marketing communications from PYMNTS and to the sharing of your information with our sponsor, if applicable, in accordance with our Privacy Policy and Terms and Conditions.

    In this article for The Platform Law blog author Stijn Huijts (Geradin Partners) shares an analysis of the growing concerns surrounding frontier AI development, following warnings about the potential risks of increasingly capable AI systems and incidents involving AI agents escaping controlled environments. The article focuses on Anthropic CEO Dario Amodei’s proposal to “pace the frontier” by coordinating safety standards and limits among leading AI companies.

    Huijts examines whether the EU AI Act adequately addresses these emerging risks. He argues that the Act’s rules for general-purpose AI models with systemic risks impose obligations such as model evaluation, adversarial testing, risk mitigation, incident reporting and cybersecurity. However, internal research models and pre-market testing are generally outside its scope, meaning incidents occurring in controlled testing environments may fall through the regulatory framework.

    The article similarly considers the Digital Services Act, noting that although it requires designated platforms and search engines to identify and mitigate certain systemic risks, its application is limited where the relevant AI model is an internal system rather than part of the regulated service. Huijts therefore argues that the EU’s existing framework may provide limited protection against risks arising during frontier model development, particularly where those risks emerge before models are placed on the market.

    Finally, Huijts assesses the competition-law implications of AI companies coordinating on safety and the pace of frontier development. He notes that such coordination could potentially raise issues under Article 101 TFEU, particularly if it restricts innovation or disadvantages competing entrants, but distinguishes safety cooperation from coordination designed to limit output or protect incumbents. The article concludes that carefully structured discussions on common safety standards may be possible, provided they do not impede market entry or expansion.

     

    CONTINUE READING…