The Case for Large Language Models in Contract Generation

Abstract

Does a contract need a mind behind it? Eliza Mik’s influential working paper says yes and concludes that large language models, lacking genuine understanding, can produce only “contractual deepfakes.” This article argues that Mik’s technical premise is accurate but her jurisprudential conclusions do not follow from it: contract law has never required epistemically validated drafting, only functionally adequate text. Through systematic doctrinal analysis, five structural defects are identified in Mik’s argument, including concessions within her own paper that are fatally inconsistent with her conclusions. In their place, this article proposes the Functional Adequacy Threshold, i.e., a context-calibrated standard grounded in the objective theory of contract and attentive to the access to justice dimension that Mik’s analysis entirely ignores. The argument is not that AI contract generation is risk-free. It is that risk calls for governance, not prohibition.