Seven withholding-tax traps hiding in modern treaties

The treaty landscape is more dangerous than it looks

Withholding-tax treaty work has always sat on a thin line between mechanical and treacherous. In 2026, that line is thinner. The Multilateral Instrument (MLI) has now amended the substantive provisions of more than 1,800 bilateral treaties; domestic courts have begun to interpret the principal-purpose test (PPT) in ways that diverge from the OECD commentary; and a growing number of treaty partners are reading beneficial-ownership requirements narrowly enough to deny relief on facts that would have qualified five years ago. The cumulative effect, in our experience advising on roughly 200 cross-border payment streams over the past eighteen months, is that the average mid-market group is leaving between 5 and 10 percentage points of effective tax rate on the table — not through structural failure, but through traps that have been quietly added to the network.

What follows are the seven traps that have cost our clients most. Each is widespread; each is fixable; none is obvious from the face of the treaty.

1. Most-favoured-nation (MFN) clauses that auto-update

A handful of older treaties contain MFN clauses that automatically apply the more favourable terms of any subsequent treaty the contracting state signs with a third party. India's treaty with Slovenia, for example, contained an MFN clause that — when Slovenia subsequently joined the OECD — was argued to import the lower withholding rates from India's OECD-counterparty treaties.

The trap is that MFN clauses are not always self-executing. Indian Supreme Court jurisprudence (notably the 2023 Nestlé decision) clarified that MFN benefits do not flow automatically: a notification by the Indian executive is required to give effect to the lower rate. Groups that had claimed MFN benefits without the notification have had relief denied retrospectively, with interest and penalty exposure. The flip side is that MFN clauses that have been properly notified are routinely overlooked by groups defaulting to the headline rate. We see this trap cut both ways: claims wrongly made, and claims wrongly missed.

2. Principal-purpose test (PPT) carve-outs

The PPT, introduced into most modern treaties via the MLI, denies treaty benefits where obtaining the benefit was one of the principal purposes of the arrangement, unless granting the benefit would be in accordance with the object and purpose of the treaty. This is a vague standard, deliberately so, and domestic courts are interpreting it inconsistently.

Three jurisdictions have produced PPT decisions in 2024–2025 that are now standard reading: the Indian Income Tax Appellate Tribunal's Tiger Globaldecision (denying treaty access to a Mauritius holding); the Dutch Hoge Raad's 2024 ruling on a Luxembourg holding (granting treaty access despite a structuring challenge); and the Spanish Audiencia Nacional's denial of treaty relief to a Maltese intermediary on a fact pattern that would have been routine in 2018.

The practical implication: the PPT is now binding on most modern treaty positions, and the analysis is fact-specific. Treaty-residence certificates are necessary but no longer sufficient. Groups using intermediate holdings should expect to demonstrate (a) commercial purpose independent of the treaty benefit, (b) a substantive presence at the intermediate level, and (c) a coherent narrative for why the structure exists in its current form.

3. Beneficial-ownership reversals

Beneficial-ownership requirements have always been a soft check on treaty access. In 2024–2026, they have hardened considerably. The Brazilian Superior Tribunal de Justiça issued a 2024 decision denying beneficial-ownership status to a Spanish holding on the basis that it had no power to use or enjoy the income; the Argentine Federal Tax Court reached a similar conclusion in 2025 for a Dutch cooperative; and the Indian tribunals have continued to narrow the concept across multiple decisions.

The pattern across these decisions is consistent: where the recipient is contractually or commercially obliged to pass the income on to a different party, beneficial ownership is denied. Back-to-back arrangements, conduit structures, and cooperatives with formulaic distribution policies are all being scrutinized on this basis. Withholding agents in the source jurisdiction are increasingly demanding contractual and operational evidence of beneficial ownership before applying the treaty rate, and the burden of proof has shifted in practice.

4. Limitation on Benefits (LoB) tightening under MLI

The MLI offers contracting states two anti-abuse options: a simplified LoB (mechanical residency-based tests) and the PPT (subjective purpose test). Most jurisdictions have adopted the PPT only; some — notably the United States, India, and a handful of Latin American counterparts — have adopted both. Where the simplified LoB applies, treaty access is contingent on the resident meeting one of several specific tests: publicly traded, ownership/base erosion, derivative benefits, or active business.

The trap is that the LoB tests are mechanical and unforgiving. A group that previously qualified under the derivative-benefits test by virtue of its ultimate ownership chain can fail it after a routine corporate restructuring — and the failure is often not identified until withholding tax is denied at source. We recommend annual LoB testing as a standard governance task for any group with material treaty-dependent income flows.

5. Royalty re-characterization in software and cloud

The classification of software and cloud-services payments as royalties (versus business profits) remains jurisdictionally fragmented. The OECD commentary draws a line at the right to commercially exploit the IP versus mere use; many jurisdictions, particularly in South Asia and Latin America, draw the line further toward royalty characterization.

The 2021 Indian Supreme Court Engineering Analysisdecision, which limited royalty characterization for software, has been narrowed in practice by subsequent tribunal decisions — particularly for SaaS and cloud arrangements. Brazil's 2024 reform similarly broadened royalty treatment for cross-border technology services. Groups receiving software and cloud revenue from these jurisdictions should expect withholding at the royalty rate unless they can demonstrate a clear contractual and commercial position consistent with business-profits classification.

6. Fees for technical services (FTS)

Many treaties — particularly Indian, Pakistani, Brazilian, and several African counterparts — contain a separate FTS article that allows source-state withholding on technical, managerial or consultancy services even where there is no permanent establishment. The rates are typically 10–15%.

Two traps are common. First, the “make available” requirement in some treaties (notably the India–US, India–UK and India–Singapore treaties) restricts FTS to services that transfer skill, knowledge or expertise to the recipient. Where the requirement applies, ordinary outsourced services should not attract FTS withholding; many withholding agents apply it nonetheless and the payee must claim a refund. The cash cost of an unrecoverable refund position is real. Second, the FTS article and the royalty article overlap on certain hybrid services; sequential application has caused double-counting in source-state assessments. Coordinating classification across the FTS, royalty and business-profits articles is essential.

7. Service PE and force-of-attraction rules

The seventh trap is structural rather than transactional. A growing number of treaties include service permanent establishment provisions that create a PE in the source state if employees are present for more than a threshold number of days (typically 183 in any 12-month period). Once a service PE is established, the source state can attribute profits to it — and a small number of treaties (notably Brazil, India, and several African counterparts) include “force of attraction” rules that pull broader categories of source-state income into the PE's tax base, even where the income was not directly earned by the PE.

With remote and hybrid work patterns now embedded across most multinational workforces, the day-count thresholds are easier to breach than they were in 2019. Groups that have not refreshed their travel-tracking and PE-risk monitoring processes since 2019 are routinely accumulating PE risk they are not yet measuring. We recommend a baseline audit of cross-border employee days for any group with material activity in PE-risk treaties — and an automated monitoring rule for any single individual approaching the threshold.

The bottom line

Withholding-tax treaty work is no longer a matter of looking up the headline rate. The MLI, the PPT, jurisdictional beneficial-ownership reversals, and the tightening of LoB, royalty and FTS interpretations have compounded into a landscape where the difference between compliance and overpayment is consistently 5–10 ETR points across a typical mid-market payment portfolio. The traps are knowable; missing them is not a research problem, it is an infrastructure problem. Build the monitoring, run the analysis annually, and refresh your treaty-residence positions before each major payment cycle. The cost of doing it is a fraction of the cost of not.