Home Artificial Intelligence Artificial Intelligence And The U.N.’s Global Tax Reform

Artificial Intelligence And The U.N.’s Global Tax Reform

Robert Goulder of Tax Notes and Lyla Latif of Warwick Law School discuss the importance of considering artificial intelligence as the U.N. aims to create a more inclusive, equitable, and effective global tax system.

This transcript has been edited for length and clarity.

Robert Goulder: Hello, I’m Bob Goulder, contributing editor with Tax Notes. Welcome to In the Pages, the video series where we take a closer look at content from our print and online publications. Our topic this week is a very timely one: artificial intelligence and tax reform.

Now, as you may know, the United Nations approved a resolution last December that could pave the way for a historic multilateral agreement on the allocation of global corporate taxing rights. The question we have is whether the United Nations will consider recent advancements in AI in their solutions. If they don’t, they could be making a big mistake. That is the conclusion of our featured article, which is titled, “Why the U.N. Must Put AI and Data on the Tax Agenda.”

To help us sort through these issues, we are joined by the author Lyla Latif, Ph.D. She is a lawyer and tax specialist with Kenya’s Ministry of Information, Communication and The Digital Economy. She’s also a research fellow at the University of Warwick’s Law School in Coventry, England. Lyla, welcome to In the Pages.

Lyla Latif: Thank you so much. It’s exciting to be here with you, Robert.

Robert Goulder: Really a fantastic article. Thank you for publishing it with us. A very timely topic. There is this expectation that going forward into the future, business models across the global economy will be shaped by AI. In fact, this is not some future thing; it’s already happening now. It’s been happening for a while. I’m curious about the challenges that this presents to policymakers in the tax area. We have these norms, these standards. We sometimes refer to them loosely, broadly, as the “international consensus.” Are they up to the task of AI? Or are they lacking something?

Lyla Latif: That’s a spot-on question, to be honest, because I think we’re grappling with that going forward, guys from Global North, guys from Global South, and we’re like, “Where do we stand?” Especially where AI now comes into play. You’re right, indeed; there’s this rapid advancement in AI that are reshaping business models across the entire economy, across the world, and this definitely presents significant challenges to international tax policy as we know it.

In my opinion, there is no policy direction, and there are no legal norms to give us guidance on how to deal with these unique characteristics of AI-driven businesses insofar as tax in itself is concerned. In my view, I think that there are three key features of AI that remain open-ended in terms of regulation that could be meaningful for tax purposes.

There is the fact that AI relies on data and that AI-driven businesses heavily depend on the collection, processing, and analysis of vast amounts of data. This data in itself is gathered often from users across multiple jurisdictions, which makes it many a times difficult to determine the value that is created by the data and to attribute it to a specific country for taxation purposes. The international or the current tax norms as we have them do not really have clear provisions on how to assign value to data and then allocate taxing rights accordingly.

There is also the fact that the development and deployment of AI systems themselves often involve multiple stages. We have data collection, and then we’ll have algorithm training, and then we have the app development. These stages themselves, they occur in different jurisdictions, which leads to the most important thing here is that fragmented value chain. While it may seem that there is nothing to tax at the collection, training, and development stages, these activities can have a significant impact on revenue and value creation, which should be considered for tax purposes. You see, once data is collected, it is used to train AI algorithms which can then be deployed to generate value for the business.

The training process often involves significant investments in computing power, human capital, as well as intellectual property. These investments themselves can lead to the creation of valuable intangible assets such as proprietary algorithms and models, which can then be licensed and sold to generate revenue. The current tax norms, as you put it to me, which focuses on physical presence and traditional value creation, I think it really struggles to capture this complexity and fairly assign taxing rights to countries involved in the AI value chain.

My last point on relating to your question on how AI presents challenges to international tax policy is that AI-driven businesses rely heavily on intangible assets. Think of algorithms, software, IP. These assets are often the key drivers of value creation in the AI economy. The current international tax norms are primarily designed for tangible assets, and they may not adequately address the challenges associated with the valuation and taxation of intangibles. This in itself can lead to misallocation of profits and the erosion of tax bases.

Even when you look at the OECD’s BEPS project, it primarily focuses on traditional business models and it may not fully capture the complexities of the AI economy. For instance, the current rules — they struggle to determine the value creation and profit allocation in AI-driven businesses as the development and deployment of AI often involve multiple jurisdictions, which makes it very challenging to assign taxing rights fairly within the pillar 1 approach.

Robert Goulder: Your paper uses this wonderful phrase, a term of language that I really appreciated. You use the term “algorithmic colonization,” which is a wonderful phrase, a powerful term in what it conveys about the long-term ripple effects of AI. We can’t anticipate there’s going to be these concentrations of economic power specifically due to AI.

Now, from the standpoint of a developed country like say the G7 or the G20, this is going to raise concerns about their tax base. But who knows, maybe it plays to their favor because the rules were written at a time when we didn’t anticipate any of this. Your paper emphasizes that these issues about tax base and responding to the misalignment, they’re particularly acute for the Global South. Can you elaborate a little bit more on that?

Lyla Latif: I like that word as well, “algorithmic colonization.” It’s actually a term that I borrow from the growing scholarship out of the Global South that rightly describes the potential consequences of AI’s uneven development and deployment across the globe.

I want you to imagine a world where you have this handful of big tech companies from the developed countries and they hold all the power in the AI game. The other one’s calling the shots. They’re concentrating economic might in their own hands. So sure, it’s a concern for tax bases of developed economies, but the real punch in the gut is felt by the Global South.

Let me give you an example of Microsoft. I was reading this the other day. It has developed a product called FarmBeats, which is essentially a cloud-based platform, and it’s designed for the agricultural sector. Microsoft is currently testing it out in developing countries, and this FarmBeats, it sort of provides farmers with access to advanced technologies like drones and sensors that can help them monitor things like soil moisture levels. Here’s where it gets interesting.

Microsoft is teaming up with other companies that are working on cutting-edge agricultural tech, such as self-driving tractors and drones that can spray pesticides. One of these partners that Microsoft is working with is called Climate Edge, and this is a company that actually describes itself as a big data broker for agricultural industry in developing nations. Essentially what they do is that they gather data from small scale farmers through various sources like NGOs, agricultural consultants, and other companies that use their platform. They then package this data and sell it to a wide range of interested parties. Think of insurance firms, certification bodies, pesticide manufacturers, major food corporations like Unilever. This to me is clearly indicative of concentration of economic power and commercialization of agriculture data by multinational corporations. This is what algorithmic colonization tries to emphasize, that having this sort of a business model by Microsoft is what is undermining the autonomy and economic potential of the Global South.

You’ve got small farmers who are the backbone of many developing countries. They’re reduced to mere data points. Their valuable knowledge and their insights are siphoned off and sold to the highest bidder. Then you have the profits from this data exploitation that flow back to the developed world while the Global South is left with crumbs. It’s like a digital form of colonialism where the rich keep getting richer and the poor are left struggling to assert their rights and capture this fair share of value that is created within their own borders.

I actually thought that the word algorithmic colonization is really befitting to actually see what’s really happening on the agenda today, so tax becomes really critical.

Robert Goulder: That’s a perfect description of it. Again, very powerful. Now let’s talk about some solutions. We’re going to fix it. Well, we’re not going to fix this. The United Nations hopefully is going to fix this. The thesis of your article, if I can be so bold as to summarize it in just a bullet point, is that the U.N. ad hoc committee, they must develop new standards and norms specifically tailored to the issues related to AI. I’d like you to give us an example here. You mentioned one in your paper. It’s about the PEs, permanent establishments.

All of our readers of Tax Notes, and the vast majority of the listeners of this are going to be familiar with the PE concept. It’s in thousands, many thousands of tax treaties. It’s in the U.S. model treaty, it’s in the OECD model treaty. It’s been around for roughly 100 years. We’ve been living under this concept. If there’s a PE, the country can tax. If there’s no PE, sorry, you’re very limited in what you can do. You specifically discuss the concept of an AI-enabled PE. That’s exciting. Can you tell us a little bit more about that?

Lyla Latif: Yes, definitely. I’m a researcher and I like to think about new ideas, and I don’t like to follow the thinking that’s already out there because I believe that we’re in a generation where we need to think for ourselves and come up with important themes that resonate with us. AI-enabled permanent establishment is something that came up. Why I thought about that is that first, you know that we can’t just rely on the existing international tax framework because it wasn’t built with AI in mind. You also said this earlier on.

When it comes to thinking about what might these new rules look like, well, one idea that I thought is, let’s think about introducing a concept called AI-enabled permanent establishment. It is my way of trying to figure out how do we — pillar 1 has had its problems with coming up with — how do we establish the nexus for digitalized companies? Do we go for the significant economic presence approach, which is limiting?

When we’re looking at AI, and I mentioned that AI is all about data collection, algorithmic training, and app development, these are different things that have to be considered from a different perspective. What should a permanent establishment look like for these AI companies? Like you said, we all know about PE. It’s essentially a threshold that determines whether a company has a taxable presence in a particular country. Traditionally, it’s just based on physical factors like do you have an office, do you have employees?

But the AI economy businesses, it’s a bit different. AI generates value in a country without physical presence by simply collecting data or even deploying AI systems. That’s where the idea of an AI-enabled PE comes in. Under this approach, in the way I was trying to conceptualize it, is that if an AI system is collecting data, it’s generating insights or it’s creating value within a country, it should be considered as a permanent establishment, even if the company doesn’t have a physical office or employees there. I really think this would be a game changer for countries in the Global South, which often serve as a source of valuable data for AI development, but they struggle themselves to capture a fair share of profits that are generated by these companies.

With the concept of an AI-enabled PE, I believe that African countries, other developing countries, could actually get to assert their right to tax the profits that are generated by AI activities within their borders. This would ensure a more equitable distribution of the profits or even the benefits of AI. That just happens to be one piece of the puzzle. We also need to rethink the way we in itself allocate profits in the AI economy. The current rules under the BEPS project, for example, they’re based on the idea that profits should be allocated to countries where value is created.

But in the AI context, value creation, like I said earlier, is often fragmented across multiple countries with data collection, algorithm development, and deployment, which happens in real time, but across different, multiple jurisdictions. The concept of this AI-enabled PE is my approach to understanding the unique challenges of taxing the value that is created by AI activities in the global economy. One thing I want to say is that even when the OECD’s pillar 1 proposal, it has a concept of significant economic presence and that also aims to tackle the issues of taxing businesses that operate digitally without a physical presence. The idea may be, well, the concept that I’m proposing, AI-enabled PE doesn’t make sense because pillar 1 was already addressing it. I think there’s a difference here, because I think pillar 1 does not fully capture the specific characteristics of the AI value chain. An AI-enabled PE would establish a new nexus threshold tailored to AI activities that focus on my three things: data collection, algorithmic development, and the AI system deployment. This approach in itself, it requires new profit allocation rules that account for the value of intangible assets like data, you can’t put a price on it, and even algorithms, which are critical drivers of value in the AI economy.

Implementing the concept of this AI-enabled PE could potentially, and maybe the ad hoc committee needs to think about this, it could potentially be achieved through bilateral or even regional agreements or changes to domestic tax laws like a unilateral measure. It can be offered as a complementary approach to pillar 1 and also the significant economic presence initiatives that are there like in Nigeria, for example, in Kenya, to ensure fair taxation of AI-generated profits. So I really don’t have all the answers. I only have ideas to invite debate on.

Oh, yes, and another thing, sorry. As you know, pillar 1 is aimed at addressing the tax challenges arising from the digitalization of the company. Now that approach only seeks to allocate a portion of the profits of large multinational corporations to the countries where the users are located based on certain thresholds and criteria. While this pillar 1 is a step in the right direction, it has some limitations when it comes to addressing the specific challenges that are posed by the AI economy. That’s the reason why I really feel that we need to distinguish that yes, pillar 1, it may not go through. If it doesn’t go through, what’s there for these AI models that are there generating profits which are being left untaxed? How can countries come into tap that sort of value from them? I thought an AI-enabled PE would make some decent sense.

Robert Goulder: A modern solution for modern times. When you put it like that, who can say no?

Lyla Latif: Yeah.

Robert Goulder: Well, that’s all the time we have. Again, just to review, the author’s name is Lyla Latif. The article is titled, “Why the U.N. Must Put AI and Data on the Tax Agenda.” You can find it in Tax Notes. Lyla, thank you so much.

Lyla Latif: Thank you so much for having me. This is great.



Denial of responsibility! TechCodex is an automatic aggregator of Global media. In each content, the hyperlink to the primary source is specified. All trademarks belong to their rightful owners, and all materials to their authors. For any complaint, please reach us at – [email protected]. We will take necessary action within 24 hours.
DMCA compliant image

Leave a Comment