This paper investigates a fundamental challenge at the intersection of fiscal law, legal theory, and artificial intelligence: the conditions under which a network of fully autonomous AI agents—having generated, reproduced, and delegated themselves without traceable human authorship—executes a commercially valid business sale and deposits cryptocurrency proceeds into an unattended non-custodial wallet. We argue that this scenario, termed the Fully Autonomous AI Transaction (FAAT), constitutes a structural trilemma for modern tax law: the simultaneous failure of personhood-based, residence-based, and self-assessment-based jurisdictional mechanisms. Drawing on primary scholarship from Avi-Yonah et al. (2025), LoPucki (2018), Hadfield & Koh (2025), and regulatory frameworks including OECD CARF (2024) and EU DAC8 (2023), we develop a three-part taxonomy of the problem and propose two novel frameworks: Transaction-Layer Fiscal Attribution (TLFA) and the Benefit-Based Attribution Principle (BBAP). The paper then subjects both frameworks to systematic critical scrutiny—five identified objections, stated in their strongest form—and strengthens the argument in response. The revised position is more modest but more defensible: the FAAT is a structural possibility that exposes genuine weaknesses in existing doctrine, and adequate response requires deliberate architectural choices that have not yet been made.
Meatless Labs (Fri,) studied this question.