Paul L. Caron

Monday, April 22, 2024

Avi-Yonah & Salama: Taxation Of Autonomous Artificial Intelligence

Reuven S. Avi-Yonah (Michigan; Google Scholar) & Lucas Salama (Michigan), Taxation of Autonomous Artificial Intelligence: Socially Sustainable Expansion of Automation and Impacts on International Tax:

This paper investigates if artificial intelligence should be taxed independently from its controllers or owners and how this could be structured and used to benefit tax administration while being a positive influence for private sector stakeholders. The main question that will be investigated is whether the reasons for taxing such entities outweigh the respective negative consequences. We propose that autonomous A.I. has started a transformation in the way legal systems around the world assign rights and obligations, and creating a tax on the profits generated by autonomous systems is not only coherent with the current business entity model of taxation, it is also an effective way to address the international tax challenges arising from A.I. operating in multiple jurisdictions and in providing a reliable structure for regulation, and potentially creating a way to safeguard socially responsible expansion in the use of automation.

The increasing relevance of robotics and artificial intelligence in modern society has generated various debates about those  technological trends. Topics ranging from liability to personhood and subjection (of these technologies) to tax have evolved from early discussions about the legal consequences of a possible and at the time distant future to concrete debates about what changes automation ought to usher upon the legal system. In tax, this has meant numerous proposals to impose a new tax that in reality collects revenue from the human owners to address an excessive subsidy of capital expenditures and the negative externalities of displacing human workers. The latter presents two issues, the first one being a decrease in tax collection from employment and the second an increase in social security costs that will be incurred by governments to deal with displaced workers.

These “Robot taxes” are inadequate to deal with the challenges of autonomous artificial intelligence that is not fully under the control of its owners, programmers or users. In the international tax arena, allocating taxing jurisdiction can become a problem if no specific measures are adopted because of the peculiar structure of the operations: while traditional multinational corporations tend to have more easily determinable allocation points for income, an AI would have its programmers be anywhere (possibly in multiple locations), the servers on which the system runs could be in one country, and the users scattered across even more different countries. Unlike in a traditional structure, location of management, payroll and assets would be of little consequence, as their role in the profits generated are minor. We thus propose that the proper jurisdiction to impose a tax on the A.I.’s profits is where the users are located.

Imposing a tax on the A.I. as an independent taxpayer with the user jurisdiction being entitled to the respective revenue addresses these issues. The current literature’s shortcomings to such a model disregard the use of an A.I. tax as a regulatory instrument and focus on analyzing the proposed policy from the ability to pay perspective or sheer analogy of the artificial systems to a corporate entity. This research therefore aims to clarify the benefits of using taxation of autonomous A.I. as a regulatory tool, whilst identifying its potential shortcomings and exploring the effects of this consideration on the existing analyses on the adoption of this policy from the perspective of overly incentivizing the use of capital over labor and the negative externalities associated with the displacement of human labor.

Scholarship, Tax, Tax Daily, Tax Scholarship | Permalink