AI Innovation Competition as a Discovery Procedure - The Role and Limits of Competition Law (Max Planck Institute for Innovation & Competition Research Paper, No. 25-20), 2025, 39
Strategic partnership agreements between digital incumbents and smaller AI developers have raised numerous competition law concerns. While little is known about the specific terms of such agreements, and enforcement remains limited, the question of how competition law should respond remains open. This paper addresses the implications of these agreements for dynamic competition, particularly regarding their impact on the conditions of access to AI models developed with critical inputs provided by big tech firms. Specifically, we address the question of whether conduct that diminishes opensource (OS) licensing of AI models should be regarded as a competition law infringement. First, we show that the OS release of AI models cannot be equated with greater innovation, including for the purposes of competition law assessment. Next, as a contribution to the normative debate, we propose grounding the competition law approach in the notion of innovation competition as a discovery procedure. In light of this theoretical framework, the chapter examines decisions in Microsoft/OpenAI and Microsoft/Mistral AI cases and offers insights into how competition law should be applied. Having identified the shortcomings of the traditional competition law framework, the analysis highlights the need for a reform of the Digital Markets Act and even the adoption of a new competition law tool that would account for the importance of maintaining and promoting free choice and access in digital markets. In particular, we recommend aligning competition policy with the protection of innovation competitionan approach that safeguards AI developers' freedom to choose their preferred licensing model while protecting against undue restrictions in cooperation agreements that may undermine dependent companies' freedom to design and pursue their own innovation strategies.
