AI: Stare Non Decisis
John McClellan Marshall
ABSTRACT
AI: Stare non Decisis identifies a new danger arising from the use of artificial intelligence in legal practice. The problem is not simply that large language models may generate fictitious citations or hallucinated authorities. The deeper concern is that lawyers, judges, and institutions may gradually delegate the responsibility for verification to machines. If unverified machine-generated authority enters the legal record, the integrity of precedent itself may be undermined. This article argues that stare non decisis is more than a legal problem. It is an intellectual example that reveals a broader crisis of verification affecting universities, hospitals, and other institutions of knowledge. AI-generated legal citations, student essays, and diagnostic recommendations all expose a common challenge: the production of plausible information is becoming easier than the verification of its reliability. The central question of AI in the modern era is therefore shifting from who governs institutions to who governs the infrastructures that organize knowing itself. The article concludes
that preserving the Enlightenment commitment to sapere aude may well require the redesign of legal and educational institutions around verification, accountability, transparency, and stewardship. The future challenge is not whether AI participates in knowledge production, but whether present-day human institutions retain the capability for responsibly determining what counts as trustworthy knowledge.
