TaxProf Blog

Paul L. Caron
Dean





Friday, June 20, 2025

Weekly SSRN Tax Article Review And Roundup: Speck Reviews Taxing Electronic Agency By Elkins & Eyal

This week, Sloan Speck (Colorado; Google Scholar) reviews a new work by David Elkins (Netanya) & Mirit Eyal (Alabama; Google Scholar), Taxing Electronic Agency. 

Sloan-speck

A defining feature of large language models and other advanced artificial intelligence (AI) systems is their opacity. Humans establish an AI’s basic framework, but enormous quantities of data drive iterative processes that ultimately determine the AI’s operation and output. From this perspective, advanced AI is a black box. In Taxing Electronic Agency, David Elkins and Mirit Eyal leverage this feature of contemporary AI systems to explore the problems with, and solutions for, taxing activities mediated by advanced AI. Elkins and Eyal argue that, in U.S. federal income taxation, opacity is neither new nor unique to advanced AI, and they deploy analogical reasoning to develop a cognizable and workable path forward for addressing the taxation of advanced AI.

More specifically, Elkins and Eyal propose that, to the extent that an advanced AI operates opaquely in ways that confound or complicate income taxation, the advanced AI should be treated like a trust. These types of opaque AI are neither mere tools nor sentient entities, with all of the metaphysical questions implicated by such a reductive dichotomy. Instead, Elkins and Eyal discern an intermediate category where advanced AI controls economic value that cannot be appropriated immediately by humans. For example, an advanced AI might manage a portfolio of blockchain assets for which only the AI knows the private key. Although the AI might disburse the portfolio’s proceeds to humans eventually, those humans may not know anything about (or even know of) the portfolio or the AI’s activities until disbursement. After reviewing several possible analogies, Elkins and Eyal conclude that trust taxation, which developed in response to similar problems involving opacity, offers a compelling starting point for taxing advanced AI under current law.

Elkins and Eyal note that their analogical method eschews “the more radical approach” of designing a bespoke taxing regime for advanced AI (21). While an analogical method leverages history and pedigree to reduce administrative costs and encourage fairness, such an approach may minimize any transformative aspects of the current technological moment. If today’s world is fundamentally different (and isn’t it always), then ideal policy may need to start from a truly blank slate. From a political economy perspective, however, lawmakers almost always draw on existing structures to construct new instruments, and Elkins and Eyal do important work in identifying the best analogue for policymakers looking to respond to advanced AI. Indeed, Elkins and Eyal’s proposal has ample detail on how current law would map onto opaque advanced AI, including a recommended expansion of the PFIC rules to encompass these “electronic business trusts” (36).

From my perspective, the issue with an analogical approach is less about initial accuracy than about opportunities for structural reform in the future. Elkins and Eyal are manifestly correct that an analogy to trusts is the best way to tax advanced AI today. But if advanced AI represents a genuine break with the prior economic order (or the culmination of a longer breaking process), then policymakers might reasonably hold out for a more holistic rethinking of taxation in this new context, even at the cost of a longer period of severely deficient tax law. The critical question is how close an analogical approach comes (or eventually will come) to ideal policy, as well as the extent to which analogical solutions temper policymakers’ enthusiasm for future reform. Path dependency looms large. Although difficult to tease out, these stakes should be considered in establishing—and critiquing—tax law responses to advanced AI.

Finally, Elkins and Eyal explicitly move the debate over advanced AI beyond questions about the relative tax burdens borne by capital and (human) labor. Today, these longstanding questions arise in compelling contexts ranging from post-industrial automation to the economic returns from consumer data. The appropriate taxation of capital income, however, implicates both substantive and administrative issues, with at least as much weight in the latter as the former. In the context of advanced AI, Elkins and Eyal do crucial work in emphasizing the ways that administrative solutions can bridge and mitigate substantive uncertainties. Whether we’ve reached the end of the Anthropocene means less than the longitudinal custom and practice of tax avoidance and enforcement that characterizes the history of U.S. federal income tax law. The future, in all likelihood, will continue to evolve along these lines.

Overall, Elkins and Eyal’s article is a valuable and significant contribution to emergent debates about advanced AI, as well as to more established academic conversations about the taxation of capital income. Elkins and Eyal’s analogical method reveals underappreciated dynamics in advanced AI that are relevant inside and outside of taxation. Policymakers, as well as academics in law and related fields, should attend to the analysis so ably developed by Elkins and Eyal.

Here’s the rest of this week’s SSRN Tax Roundup:

Editor's Note:  If you would like to receive a daily email with links to tax posts on TaxProf Blog, email me here.

https://taxprof.typepad.com/taxprof_blog/2025/06/weekly-ssrn-tax-article-review-and-roundup-speck-reviews-taxing-electronic-agency-by-elkins-eyal.html

Scholarship, Sloan Speck, Tax, Tax Daily, Tax Scholarship, Weekly SSRN Roundup | Permalink