AI regulation and the protection of source code

Artificial intelligence (AI) has been rapidly developing since the mid-20th century. Today, AI is deployed in many aspects of our daily lives and is increasingly ubiquitous. At the same time, concerns about AI, including discrimination, privacy and security, have prompted calls for greater regulation. To this end, regulators may seek access to AI’s source code (the human-readable program for the software). However, many trade agreements contain provisions prohibiting source code disclosure as a condition for market access in the relevant countries. This article considers the tension between AI regulation and these provisions. In situations where such provisions apply, for AI regulation to remain compliant, we argue that regulators can only require access to source code in exceptional cases where the general exceptions in the General Agreement on Trade in Services (GATS) apply.