Antitrust-compliant AI industry self-regulation

Abstract

This paper argues that an agreement among AI engineers to not engineer unsafe AI could meet the requirements for antitrust legality under the Rule of Reason, thus avoiding per se illegality. The agreement must (1) be promulgated by a professional body, (2) not directly affect price or output level, and (3) seek to correct some market failure, such as information asymmetry between professionals and their clients. Professional ethical standards promulgated by a professional body prohibiting the creation of unsafe AI could plausibly meet all three requirements. This proposal is analogous to existing canons of medical and engineering ethics that prioritize safety and competence. – AI-generated abstract.