The UN Tax Convention draft protocol on cross-border services: clear direction but crucial details left open

The much awaited Co-Lead’s draft protocol on the taxation of income from cross-border services was published on 20 July 2026. Like previously released papers of the workstream, the text likely does not reflect the views of all countries at the negotiating table and might still undergo important changes. Important questions are still left open, and there are several technical issues. However, it proposes a clear approach and should constitute a good basis for further discussions.

The 2025 UN Model Convention leaves its mark, but with some clear departures

With 21 articles, the protocol is designed as a self-standing multilateral treaty that largely follows the structure of bilateral income tax treaties (with a narrower scope, focused on services income) and much of its language is inspired by the UN Model Tax Convention, including the latest updates agreed by the UN Committee of Experts in 2025. There are some important differences though, as set out in this table below:

Here is my analysis in more detail:

Like the UN Model, the protocol allocates limited taxing rights to the country in which services income arises, while the country of residence of the income’s recipient retains the residual taxing right once it has eliminated double taxation. It also distinguishes between automated digital services, insurance premiums, and a general category of services fees. However, it carves out income from transport services from its scope.

A key departure lies in the formulation of nexus rules: under the UN Model, the country where the payer resides may tax payments for services on a gross basis at a bilaterally agreed rate, even where the provider has no physical presence there. For automated digital services, art. 12B adds a deemed net method alongside the maximum gross rate (more on this below). The Protocol’s articles broaden the nexus rules to additional circumstances. Its articles on both fees for services (general category) and automated digital services trigger source taxation not only where the payment originates from a country, but also where services are physically performed, where the consumer is resident, or where the service is deductible for tax purposes (for the general category), or where end users are located or user data is generated (for automated digital services).

A further deviation from the UN Model Convention is article 9 of the Protocol. Under the UN Model, a country may also tax a foreign enterprise’s business profits on a net basis at the domestic corporate tax rate if the enterprise has a permanent establishment (an office or other fixed place of business) on its territory. In respect to services, the UN Model also deems the foreign enterprise to have a permanent establishment if it provides services in the country through employees or other personnel for six months or longer in a year. Article 9 of the Protocol dispenses with this threshold and provides for net-based taxation at the domestic rate in two cases:

• first, where an enterprise “carries on business” in another state through employees or agents, without any time threshold: the net income to be taxed is the income that is derived by these employees or agents. How to calculate this in detail is still unspecified.
• second, where the company elects net taxation for any of the three services categories covered by the Protocol. Here, the net income is calculated based on a “reasonable allocation of [the enterprise’s] profits from the relevant business activity, taking into account the gross revenues generated in that State as compared to the gross revenues of the enterprise from such business activity”. The method is reminiscent of the simplified net option of art. 12B of the UN Model on automated digital services but is distinct in several respects: whereas art. 12B allocates 30 per cent of the company’s local revenues multiplied by the company’s profitability ratio, “reasonable allocation” leaves much room for interpretation. This may serve to accommodate variations in capacity to implement sophisticated rules but comes at the expense of certainty. The term “relevant business activity” also leaves open whether and at what level the profits of a company should be consolidated or segmented, whereas art. 12B sets a relatively clear rule in that regard.

Another key difference from the UN Model and most existing tax treaties lies in art. 2 of the Protocol, which besides income taxes also includes digital services taxes, equalisation levies, and excise taxes in its scope — and therefore under its limits.

The devil is in the gaps: relationship to existing agreements and rates

The current draft does not yet propose how it would interact with existing tax treaties. This will matter most for its practical impact. As more than 80 per cent of all cross-border services flows are currently covered by bilateral tax treaties, the aggregate impact of the protocol would be low if it only applied between countries that currently do not have a tax treaty with each other (even though it may still be high for individual countries. Low-income countries have particularly small treaty networks).[1] Moreover, if the Protocol leaves countries to decide for each individual treaty whether they would like to subject it to the protocol, the protocol might lose some of its multilateral character and may function more like another model convention. The draft does not settle the issue: it states that this would be decided only after art. 21 of the Framework Convention is discussed. When the protocol was discussed at the session back in February, this was already the most controversial question.

Another aspect currently left open concerns the maximum gross rates applicable for the different types of services. It is not clear whether the empty brackets in the text mean that rates should be negotiated bilaterally or if their negotiation is simply deferred to later in the negotiations. . In a multilateral protocol, however, a common rate agreed among all parties is equally conceivable. Nonetheless, the fact that net taxation options exist across the board might ease some of the tension around rates. Nevertheless, their level will have an impact on how often the net option will be used in practice. This affects both the administrative resources needed to apply the protocol and the pressure to further clarify the net methods.

Important technical issues need to be addressed

As might be expected at a draft stage, the protocol is not yet fully developed from a technical perspective. Beyond a lack of definitions for some of the key terms, several issues merit further discussion:

Avoiding overlaps in cases where income arises in more than one source state

As described above, income from services under the protocol can arise in the country from which the payment is made (or where it is deductible), the country where the services are performed, or the country where the consumer is located. For automated digital services, the location of end users and of user data also matters. These triggers may point to different countries: an enterprise may, for instance, render services to a client in one country, receive the underlying payments from another country, and perform them physically in a third country. Negotiators will need to address such situations, since a rule allowing overlapping tax claims will likely invite resistance. Even though the current wording suggests a hierarchy of sorts between different triggers, it is not clear that it can prevent such overlapping claims. The sourcing rules developed for the OECD’s Multilateral Convention to Implement Amount A of Pillar 1 could serve as inspiration, since they consider some specific business models where the source of a payment and the consumption of a service may often not be in the same country. Thresholds for the different nexus rules could be considered to pre-empt tensions, as well.

Making sure the net taxation option cannot be used to avoid source taxation

Article 9, which provides for net-based taxation as described above, may be vulnerable to exploitation by taxpayers. Under the current draft, enterprises may be able to minimize taxation at source by placing a few employees in a country because then only the income attributable to them (which may be a small part of all income earned from a given country) would be taxable by the source state. This is unlikely to have been the intention of the drafters, but it would be worth clarifying this further. Another weakness of article 9 is that, unlike art. 12B UN Model, it does not refer to the profits of the entire multinational enterprise group in case the company is part of one. Groups might escape taxation in a source state by only leaving little profits in the company providing the services. More generally, the protocol needs a general anti-avoidance rule, which could also help prevent treaty shopping.

Consider more generous foreign tax credit rules

In art. 10 on the elimination of double taxation, the protocol employs the language used in the Model Conventions and in many existing tax treaties. It provides for an item-by-item foreign tax credit – a relatively restrictive method that might not always fully eliminate double taxation – and implicityly leaves further calculation details to domestic law.

As our research research shows, countries often have more generous provisions in their domesic law than in treaties, allowing for instance cross-crediting of different items of income or carry forward excess credits. Although they often apply them irrespective of stricter provisions in tax treaties, it should be considered to inscribe such provisions directly in the protocol.

The protocol’s appeal remains uncertain, but there is something in it for (almost) everyone

With negotiators discussing the draft in New York  this week (10 and 11 August), we will get a first glimpse of how it is received by different countries. Beyond the answers to the open questions outlined above, the decision of a country to sign a protocol and subject its existing treaties to it likely depends on whether it is a net services importer or exporter with respect to the other participating countries, as well as the size and content of its current treaty network.

Net services importers with few tax treaties

These countries might be hesitant to sign the protocol, since – depending on the rates ultimately agreed – they might significantly limit their own taxing rights, possibly leading to less revenue collection. However, given that income from remote provision of services is one of the few types of income for which most major exporting countries do not grant a , the protocol could be of benefit even for this group of countries. Through a protocol, a country can ensure that the taxes it levies at source are credited and thereby making it less likely that their burden is passed on to local customers, without having to give up taxing rights on other types of income where there is little to gain in return.

Net services importers whose imports are largely covered by tax treaties

Here, the calculation will depend on how much existing treaties safeguard taxing rights on services. This varies significantly across countries’ networks. However, existing clauses usually only cover technical, management, and consultancy services, and a more limited nexus focused on the residence of the payor. The calculation will therefore also depend on the share of these services in the import mix and how countries expect the mix to evolve in the future.

Net services exporters

The direct incentives to sign appear smallest at first sight, especially as the United States (in relation to which many big services exporters are net importers) has left the negotiations. However, they would benefit from the inclusion of net taxation options where existing treaties only contemplate gross taxation of technical services. They can also achieve lower gross rates and greater certainty for their businesses in relation to countries where currently no tax treaty exists – provided the latter sign on and provided technical issues are ironed out. Finally, a widely accepted protocol might reduce the number of disputes that currently exist around the boundaries between business income, royalties, services, and “other income” under tax treaties.

Concluding thoughts

The draft protocol is a good foundation for further discussion. It proposes a clear direction: broader source taxing rights over services income, combined with options for net taxation. Whether countries will sign, however – and which ones – will likely depend on how open questions are resolved: the level of the rates, the relationship with existing treaties, as well as technical questions, some of which I have highlighted above.

Watch this space!

ICTD and the UN Tax Convention – more blogs by Frederik Heitmüller

Check out Frederik’s related analyses:

• Double taxation relief on remote services – implications for UN Tax Convention Protocol 1 negotiations
• UN Tax Convention Reading List (Part 1 on the UN process and institutional context, Part 2 in the taxation of cross-border services, Part 3 on dispute resolution and prevention)
• UN Tax Convention Negotiations: Where are we at and where are we headed?
• A decision on decision-making and no disputes on dispute resolution: the Organisational Session for UN tax negotiations
• No ‘winners or losers’? Reflections on the UN tax negotiations

__

[1] The following table shows aggregate shares and average shares of services imports (all categories, including transport) that are covered by tax treaties, by income group:

Source: Compiled by the author, based on data from the WTO-OECD BATIS database (year = 2023), ICTD Tax Treaty Explorer, OECD Corporate Tax Statistics