- tax clinic
- INTERNATIONAL
VAT challenges in AI product development
Related
R&D credits and the computer rental rules time forgot
The historical shift in transfer pricing penalty enforcement
The end of deferral: Calculating QOZ gain recognition on Dec. 31, 2026
Editor: Mary Van Leuven, J.D., LL.M.
Artificial intelligence (AI) is now embedded in products and services that resemble traditional software or cloud–based tools as businesses integrate AI into existing offerings. In many cases, AI is introduced as a feature enhancement or operational upgrade, not as a fundamental change to the service. But as automation deepens, the nature of the supply can shift. AI may move from a background tool to the primary source of value. Human involvement may decline. Customers may increasingly use self–service interfaces that generate results independently. At that point, a service that was never treated as digital may begin to function like one, creating value–added tax (VAT) consequences that were not part of the original analysis.
The subtle tax shift
This matters because most VAT laws around the world assign specific classifications to digital services, and those classifications determine where tax is owed. More than 100 jurisdictions require nonresident digital services providers to register for and collect VAT. Therefore, when a service transforms through the use of AI into what can be defined as a “digital service,” this shift may thus create a VAT or digital services tax obligation.
VAT exposure does not always appear with a new product launch. It can emerge gradually as AI features accumulate and reshape how a service is delivered. A VAT position that was accurate at launch may no longer reflect how the product operates. This item focuses on the deployment and commercialization phase of AI product development, where VAT issues most often arise. Through two scenarios, it highlights where businesses are most likely to encounter VAT exposure and how they can prepare.
An international sale can trigger unexpectedtax obligations
Scenario 1: BlackAI is a startup that offers a subscription–based app that uses artificial intelligence to create custom marketing graphics, thus competing against the more traditional “Madison Avenue” ad agencies. A user enters a text prompt or uploads a sketch, and the AI instantly generates a polished design. The service is fully automated, with no human designers involved. BlackAI is based in the United States, but online distribution quickly attracts customers worldwide.
The founders may not expect that monetizing the app can trigger VAT obligations outside the United States almost immediately. Many countries treat services such as BlackAI’s as digital or remotely supplied services, which under most VAT systems places tax where the customer is located. As a result, a sale to a customer in a country such as France, Singapore, or South Africa may require BlackAI to charge and remit local VAT, even without any offices or employees there.
For a company such as BlackAI, early international sales can quietly create VAT exposure that can quickly become material, given that the average standard VAT rate is around 19% (Organisation for Economic Co–operation and Development, “Consumption Tax Trends 2024“), and statutes of limitation are often extended in case of noncompliance.
Definitions and thresholds
Definitions of “digital services” and VAT registration thresholds vary across jurisdictions. Some countries impose VAT from the first transaction, while others apply low thresholds or define digital services broadly enough to capture many AI–powered offerings. Jurisdictions also use different methods to determine customer location, relying on billing address, internet protocol (IP) address, payment data, or a combination. Without systems that accurately apply these rules, even a small number of early cross–border sales can trigger unexpected VAT obligations.
BlackAI faces immediate operational challenges because it must correctly identify customer location and status, apply the appropriate VAT treatment, and issue compliant invoices, with mistakes exposing the company to penalties and audit risk. Although VAT compliance may not initially rank as a priority for AI startups, exposure can build quickly, making early attention to VAT critical to avoid costly remediation and disruptions to growth.
When a simple upgrade rewrites your tax profile
Scenario 2: TutorCo, a U.S. company, has long offered live, one–on–one online language lessons to consumers around the world. Students connect with human tutors over video, and TutorCo delivers the service from the United States. Under VAT rules, these lessons did not qualify as digital services. Tax authorities treated them as personal educational services taxed where the supplier is located. As a result, TutorCo did not charge VAT in customer countries and focused on U.S. sales tax where applicable. The company expanded globally without engaging with foreign VAT systems.It eventually launched an AI–powered learning portal. The new feature, an interactive chatbot that provides automated conversation practice and feedback, aimed to enhance the student experience and reduce delivery costs. TutorCo offered it both as a stand–alone subscription and as part of a bundle with live tutoring.
While the business sees this as a modest upgrade, the VAT implications are significant. The AI portal delivers language practice over the internet with minimal human involvement. In many jurisdictions, tax authorities are likely to treat this as a digital service. For consumer transactions, this often shifts the place of taxation to the customer’s location. A student in the United Kingdom using the AI portal may be considered to receive a taxable service in the United Kingdom, even though TutorCo remains based in the United States. This change likely brings TutorCo’s AI offering within the scope of other countries’ similar rules. Another particular challenge for TutorCo is that some jurisdictions exempt education. TutorCo may have benefited from those provisions if it met the particular conditions, but the involvement of AI could be considered a significant change in the service so that it no longer meets the specific definition of an exempt education service.
If tax authorities determine that the AI service is subject to local VAT, TutorCo must act quickly to manage compliance. The company may need to register for VAT in multiple jurisdictions, either directly or through simplification mechanisms. TutorCo must also update its invoicing systems to reflect local VAT rates and requirements. When the AI service is sold separately, invoices may need to show VAT charged at the customer’s local rate. In contrast, the live tutoring component may remain outside the scope of foreign VAT when sold alone.
When TutorCo bundles AI and human services, it must decide whether the arrangement constitutes a single bundled supply. If TutorCo sells the AI chatbot and live tutoring as one package, it may need to allocate value between the components and apply different VAT treatments. Misallocating value can cause TutorCo to over– or undercollect VAT, creating compliance risk. In some cases, tax authorities may instead treat the AI component as predominant and subject the entire bundle to VAT.
TutorCo must also distinguish between consumers and business customers. If a language school or corporate client purchases access to the AI portal and provides a valid VAT number, TutorCo may not need to charge VAT. Instead, the business customer may need to self–assess the VAT. This requires TutorCo to validate VAT numbers and issue invoices that state the customer’s responsibility to account for the tax.
TutorCo’s VAT exposure likely began when it launched the AI portal, even though the company did not enter new markets or change its legal structure. Because the shift in tax treatment was subtle, TutorCo may not have noticed it, allowing exposure to grow each month through unpaid VAT, penalties, and interest. Tax authorities could also reassess the nature of TutorCo’s overall offering if the AI feature becomes central to the customer experience, potentially reclassifying even the live tutoring component as a digital service.
TutorCo’s experience shows how a seemingly modest AI enhancement can fundamentally change the service’s VAT character. Businesses that fail to reassess their VAT position as AI becomes more central to their offerings risk discovering — often too late — that their tax treatment has shifted without warning.
Looking ahead: Where automation may take us
As AI capabilities expand, new patterns are emerging that raise questions that current VAT frameworks were not designed to address. AI systems that act on behalf of users, including tools that initiate or complete transactions autonomously, challenge traditional assumptions about how to identify the supplier, the customer, and the location of a supply.
Some AI tools may begin to intermediate between parties by matching users, facilitating exchanges, or guiding transactions from start to finish. If an arrangement starts to resemble marketplace–style intermediation, it introduces questions about whether new VAT responsibilities could arise, such as collecting VAT on underlying transactions, complying with platform reporting requirements, or potentially creating exposure to digital services tax regimes. These outcomes are not predetermined, but they illustrate the uncertainty businesses face when AI begins to coordinate or influence economic activity.
Some AI–driven models rely on user data rather than monetary payment. For example, an application may offer free access in exchange for users agreeing to allow the AI to be trained on their interactions. These arrangements raise questions about whether data constitutes consideration for VAT purposes and how such value should be measured. Infrastructure decisions, including where AI models are hosted, may also affect VAT outcomes where local rules treat server location as relevant to establishment.
At the same time, tax authorities are expanding their own use of automation. Real–time reporting systems and AI–driven analytics allow authorities to identify mismatches between how a service operates and how businesses report it for VAT purposes. As service delivery and enforcement both become more automated, businesses are likely to face new compliance expectations that existing VAT frameworks do not fully address.
Looking ahead, AI’s broader economic impact may also prompt policymakers to consider targeted forms of AI–related taxation. For instance, policymakers may be considering an AI usage tax based on the tokens consumed to compensate for losses in payroll tax or other negative externalities resulting from the use of AI. As a result, the tax system in an AI–driven world may look different from the current one.
Monitoring VAT exposure
AI is changing how services are built and delivered and, with that, how they are taxed. As automation becomes central to a product, it can shift a business into new VAT obligations. These shifts often go unnoticed until exposure has grown, and by then, the cost of catching up can be substantial. A single AI feature, even one launched quietly, can trigger VAT registration, invoicing, and reporting requirements in many countries. In some places, one sale is enough to create a legal obligation. If a business misses the moment when its service falls under digital VAT rules, it risks back taxes, penalties, and reputational damage.
VAT should not be addressed only after a product launches. It belongs in product planning, pricing, and customer experience decisions. Product and tax teams must work together from the start. As services become more automated, tax authorities expect compliance and are increasingly equipped to detect when it is missing. Automation changes not only what a business delivers but also what it is responsible for. Companies that do not reassess their VAT position as their products evolve may face risks they did not anticipate.
Editor
Mary Van Leuven, J.D., LL.M., is a director, Washington National Tax, at KPMG LLP in Washington, D.C.
For additional information about these items, contact Van Leuven at mvanleuven@kpmg.com.
Contributors are members of or associated with KPMG LLP.
The information in these articles is not intended to be “written advice concerning one or more federal tax matters” subject to the requirements of Section 10.37(a)(2) of Treasury Department Circular 230. The information contained in these articles is of a general nature and based on authorities that are subject to change. Applicability of the information to specific situations should be determined through consultation with your tax adviser. These articles represent the views of the authors only and do not necessarily represent the views or professional advice of KPMG LLP.
