AI Makes Tracking Foreign Digital Taxes Harder. Here’s What to Do

The IRS’s final regulations for foreign tax credits for digital taxes cast a broader net than just addressing treatment—they may prevent taxpayers from obtaining foreign tax credits because of artificial intelligence, which enables intellectual property to migrate seamlessly across borders.

Although the final regulations were issued two years ago, new developments in AI impact their evolution.

Implementing digital taxes can be messy, and authorities may not know whether to base taxpayer location for nexus on user location, billing address, IP address, or market participation. None of these factors by themselves reliably track where income-producing activity occurs, but it’s important to track them all.

The IRS’s Notice 2023-55 permits taxpayers to use the prior regulations when applying foreign tax credit rules to mitigate the potential impact. Taxpayers must carefully consider both sets of regulations and determine which provides a more favorable outcome.

Interplay Between Regulations

The final regulations imposed a new activities-based attribution requirement that dissolved the foundation on which traditional IP sourced its foreign tax credits. The requirement disqualifies tax on net income when expenses are incurred outside the taxing jurisdiction.

This creates a complex compliance burden by forcing granular alignment of sourcing rules on development, enhancement, maintenance, protection, and exploitation functions.

However, IRS Notice 2023-55 (extended by Notice 2023-80) suspends the disruptions of the attribution test by allowing taxpayers to rely on the pre-2022 prior regulations in their entirety. This is especially beneficial for owners of traditional IP income, which is generated from an IP product used in taxing foreign countries.

For most traditional IP entities, expenses related to the IP income occur either in the country where the IP is used or in another country. Under the prior regulations, such taxes were creditable, even when expenses occurred outside of the taxing jurisdiction, if the IP is used in the taxing jurisdiction.

But under the final regulations, taxes wouldn’t be creditable because of the attribution test requirement because expenses migrate to different countries.

Under prior regulations, for example, taxes on IP income generated by a French entity owned by a US corporation would be creditable even if its expenses migrated to Germany. However, under the attribution test, the IRS would disallow any credits.

When the IP is no longer developed in the country imposing the tax, then from a US perspective: The tax is no longer on IP activity. Instead, the tax is viewed as a tax on “market presence” or customers. In short, under the final regulations, this is considered a digital tax characterization because under the attribution test there is a lack of in-country activity.

IP income is considered “non-traditional” and “digital” when the entity receives income from an IP product not located in the taxing country. Income is sourced instead based on mere market participation or customer. Such taxes aren’t creditable under both the prior regulations and the recently issued final regulations, because absence of use of IP in the foreign country is insufficient to create nexus.

AI and Nexus

Taxpayers must understand the interplay between the final regulations and IRS Notice 2023-55, as well as how their AI development impacts nexus. Because AI is so mobile, modern AI systems rely on cross-border teams and globally sourced data. This allows development, training, and deployment activities to shift tax locations rapidly.

The heart of the issue is that when AI sends activities outside the taxing country, business owners are realizing far too late that they are no longer meeting traditional nexus rules. In practice, it is common to discover these shifts only after the traditional link between IP, income, and taxing country has already weakened.

This is important because qualifying for the credit requires a foreign tax on IP income to pass a number of nexus tests.

Let’s say that during a period in which a controlled foreign corporation incorporated in France outsources research and development IP to India, AI tools are no longer using the IP in the original country. When activities no longer exist or IP is no longer sourced in-country (France in this example), a foreign tax can fail the activities and sourcing tests.

AI-induced jurisdictional shifts can cause disallowance of foreign tax credits for IP income, because IP is no longer necessarily used in the taxing country. Such shifts can lead to failing both the activities test and the requirement that income be sourced to the place of IP use.

Foreign Tax Credits

Despite changes in the tax landscape and the development of AI, taxpayers can follow two key recommendations to strengthen the creditability of their foreign taxes.

Pick prior or final regulations. IRS Notice 2023-55 provides an election for temporary relief that will benefit most taxpayers. The election is made by filing a return based on the prior regulations. No separate election statement is required, but the regulations must be applied in their entirety. Selective application of the sourcing regulations isn’t permitted. While calculating which regulations are more beneficial will make the compliance burden more onerous, it could prompt considerable tax savings.

Document activities. Taxpayers should implement robust documentation frameworks that track where AI-related development, enhancement, maintenance, protection, and exploitation activity occurs. Useful documentation includes payroll records, cloud invoices, region settings, intercompany agreements, engineering decision matrices, change-management logs demonstrating control over model development and deployment, and any other items that provide insight into the shifting of IP use.

Avoiding Surprises

It is critical to proactively assess exposure to digital taxes triggered by customer or user location. Monitoring billing practices, platform access points, and IP characterization are essential to avoid being inadvertently subjected to non-creditable foreign taxes.

In an AI-driven economy, the potential loss of foreign tax credits must be weighed proactively and explicitly in structuring decisions rather than being discovered after audit. This will allow taxpayers to select the best strategy to claim foreign tax credits, depending on their specific global footprint.

This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law, Bloomberg Tax, and Bloomberg Government, or its owners.

Author Information

John Rose is tax director of Aprio’s professional practice group.

Write for Us: Author Guidelines