The UN Just Put AI on the Transfer Pricing Table

Start with the problem most transfer pricing practitioners are tracking the wrong forum for. When people think about how AI-delivered cross-border services will be taxed, the reflex is to look at the OECD - Pillar One, Amount B, the endless refinements to the arm's-length principle for baseline distribution and marketing functions. That's not wrong, but it's incomplete. There is a second, quieter negotiation running in parallel at the United Nations that could end up mattering just as much, and it is the one actually putting AI's name on paper right now.

Here's what changed. The UN Framework Convention on International Tax Cooperation has been negotiating, since early 2025, a framework treaty plus two 'early protocols' - one on dispute resolution, and one specifically on taxing cross-border services. The Co-Leads published a draft text of that services protocol on 20 July 2026, and during the Fifth Session of negotiations, held at UN Headquarters from 3 to 13 August 2026, multiple member states pushed to have artificial intelligence explicitly captured within its scope. That same session saw other states raise concerns about overlapping nexus claims for service fees, and a separate bloc pushing for optionality in how the protocol's provisions get applied. In other words: the machinery for deciding which country gets to tax AI-delivered services is being built right now, in a forum most Indian TP practitioners aren't reading transcripts of.

Why this matters more than it looks: India occupies an unusually exposed - and unusually powerful - position in this specific fight. On one hand, India has historically been among the loudest voices for expanding source-country taxing rights over digital and cross-border service income; the UN process exists substantially because countries like India argued the OECD's two-pillar solution didn't go far enough for market/source jurisdictions. On the other hand, India is also the world's largest base for AI-enabled global capability centres and IT/ITeS delivery - the exact category of cross-border service flow this protocol is trying to pin down. If the AI carve-out in the services protocol ends up defining nexus or taxing rights in a way that diverges from how OECD Pillar One and Amount B treat the same AI-delivered functions, Indian-headquartered groups and Indian subsidiaries of multinationals could find themselves benchmarking the same intercompany service flow against two different rulebooks depending on which counterparty jurisdiction is involved. That is not a hypothetical compliance headache; it is a structural one, because transfer pricing documentation is built around a single delineated transaction and a single most-appropriate-method choice, not a dual-track nexus test.

There is also a functional-analysis problem lurking underneath the treaty politics. Cross-border services protocols, going back to earlier UN work like Article 12B on automated digital services, tend to draw bright lines based on where a service is 'delivered' or 'consumed.' AI complicates that in the same way it complicates DEMPE analysis for intangibles: an AI system trained in one jurisdiction, fine-tuned or RAG-augmented with client data in a second, and delivering inference-based output to end customers in a third doesn't map cleanly onto any single-jurisdiction nexus concept the drafters are likely working from. If negotiators write a definition of 'AI services' into the protocol without engaging with how multi-jurisdictional AI pipelines actually function, they risk creating a nexus rule that transfer pricing professionals will spend the next decade trying to reconcile with functional reality - much the way 'significant people functions' language under the OECD's authorized approach took years of practice to operationalize.

The open question, and the one worth writing toward rather than around: does India's negotiating position on this protocol actually reflect an analysis of how Indian GCCs and IT exporters would be affected if AI services get a distinct, UN-defined nexus test that differs from OECD treatment - or is India's source-country advocacy here running on inertia from an earlier era of BPO and call-centre economics, before AI-native delivery models existed? That's not a rhetorical question so much as a genuine gap in the public record; the UN's own tracking shows the draft protocol text was only published in July 2026 and is still being contested clause by clause. A practitioner with both technical AI fluency and TP grounding is unusually well positioned to make that case publicly before the text hardens - which may be the real opportunity here, separate from whatever the final protocol says.

 

Who Performed the Function?

A multinational files a patent for a new compound. Tax authorities in two countries want to know which entity in the group is entitled to the royalty stream the patent will eventually generate. The test they apply is old and, until recently, reasonably reliable: find out who performed the Development, Enhancement, Maintenance, Protection and Exploitation functions for the intangible - DEMPE, in transfer pricing shorthand - because whoever bears the risk and does the deciding gets the residual profit. For decades this worked because you could point to a building: a lab, a notebook, a committee that signed off on which candidate molecule to pursue. Increasingly, the person you would point to typed a prompt, set an agent running overnight, and reviewed a shortlist of candidates over coffee the next morning. The system generated the options, ran the simulations, in some cases proposed the formulation that went into the filing. Who, exactly, performed the function? 

This is not a hypothetical irritating a handful of transfer pricing specialists. An Australian tax advisory has recently flagged what it calls a DEMPE-shaped hole - cases where AI itself is doing the value-creating work and the existing test simply has no place to put that. A parallel problem is surfacing in patent law, where AI-generated inventions are quietly breaking the doctrine of inventorship the same way they are breaking the tax doctrine of significant people functions. Two different bodies of law, built for different purposes, are running into the identical wall: both assume that value is created by an identifiable human making an identifiable decision, and both are discovering that the human's contribution has thinned to something closer to selection and approval. 

 The instinct is to treat this as a compliance problem for corporate tax departments. It is that. It is also a preview of a question that is going to reach every individual who has taken the advice - now common, and not wrong - to stop being a mere user of AI and start being a builder with it. The advice is sound as far as it goes. But it assumes the world has a stable way of recognising, after the fact, that a given piece of work was built rather than merely retrieved. Increasingly, it does not. If the design decision, the judgment call, the actual creative leap happened inside the model, and the human's visible contribution is a prompt and a sign-off, then whatever institution eventually has to decide who gets the credit - a patent office, a performance review, a promotion committee, a funding panel, or a tax auditor deciding which entity in a group actually earned a royalty - is going to ask the same question the DEMPE test is asking of multinationals right now. Where, precisely, did you sit in this? 

It gets sharper still. Some of the institutions asking that question are, at the same moment, handing the question itself to AI. A peer-reviewed study on agentic transfer pricing work found it running nearly a hundred times faster than a human analyst, and wrong in roughly one case out of five or six. That is the position builders are actually in: producing work at a pace where the artifact looks finished, inside a system that is simultaneously trying to verify whether the artifact - or the reasoning behind it - was genuinely someone's, at a speed that does not allow careful verification either. Credit and reliability are being tested at the same moment, by the same overstretched apparatus, and neither test is mature. 

None of this means the advice to build rather than merely search was wrong. It means the second half of that advice has arrived earlier than expected: it is not enough to have used AI to produce something. What will be asked, by whoever is deciding whether the work was yours, is narrower and harder - what did you decide that the model could not have decided on its own; what judgment did you make that the output alone does not show; what risk did you actually carry if the decision turned out to be wrong. That residue is what DEMPE is trying, clumsily, to locate inside multinational R&D right now, and it is what will eventually be asked of anyone whose work passed through an agent before it reached a reader, a patent examiner, or a manager. 

The practical instruction for builders is not to use AI less. It is to keep a different kind of record than most people currently bother with - not a log of what the model produced, but a trace of the decision points where a different judgment would have produced a different outcome, and where that judgment was demonstrably yours. Tax authorities are already learning, case by case, that the old test - find the person who performed the function - cannot be answered by pointing at a finished patent or a tidy set of transfer pricing documentation. It has to be answered by reconstructing a decision trail. Builders who have not been keeping one are going to discover, the way multinationals currently are, that the absence of a trail is itself the answer to the question of who performed the function - and it is rarely the answer they wanted.

Three Days to Prove Arm's Length

Start with the fact pattern, because it's more revealing than any policy paper. A Lloyd's India-regulated insurance entity received a TPO show-cause notice giving it three days to produce documentary support for its transfer pricing position. It didn't manage it in time - not because the evidence didn't exist, but because its compliance bandwidth had gone into IRDAI regulatory requirements rather than TP file-building. By the time of the appeal, the taxpayer had assembled contemporaneous emails, contracts and invoices that plausibly supported its position. The ITAT Mumbai admitted this evidence under Rule 29 and sent the matter back for fresh adjudication. On the surface this is a routine remand order. Read against the backdrop of where Indian tax administration is heading, it's a warning shot.

What's changed is the operating environment around these notices. CBDT has been public about using AI-driven risk analytics - NUDGE, Project Insight, INTRAC - to flag taxpayers and compress the distance between detection and action; the department has described the next phase of AI deployment as 'more intense.' Separately, the Income-tax Act, 2025 and the Income-tax Rules, 2026 were framed explicitly as ushering in an algorithm-based, technology-driven assessment ecosystem, with the CBDT Chairman describing simplified statutory language as something that will 'enable algorithm-based implementation.' None of this is inherently bad - faster detection and cleaner drafting are good things. But faster detection paired with unchanged (or shortened) response windows means the gap between 'flagged by the algorithm' and 'must produce five years of DEMPE-level documentation' keeps narrowing, while the underlying task - assembling functional, contractual and economic evidence for related-party transactions - hasn't gotten any easier or faster to do properly.

Here is the piece a generic Taxmann write-up on this ruling will miss: India has, in parallel, been building an entire vocabulary for governing AI systems responsibly. The Finance Ministry has laid out how RBI's FREE-AI framework and MeitY's India AI Governance Guidelines are meant to apply to financial-sector AI, organised around principles like 'Accountability,' 'Understandable by Design,' and 'Trust is the Foundation.' These are aimed at banks and NBFCs deploying AI in lending and risk models. But the tax department's own algorithmic risk-scoring - the system that generates the show-cause notices and compressed timelines that produced this ITAT remand - sits outside that governance conversation entirely. Nobody is asking whether NUDGE or INTRAC's flagging logic is 'understandable by design' to the taxpayer who receives a three-day notice, or who is 'accountable' when an automated flag compresses due process past the point where a genuine, good-faith taxpayer can respond. The state has written AI governance principles for everyone except itself.

The CBDT's own APA numbers hint at how sophisticated taxpayers are already responding to this asymmetry: a record 220 APAs signed in FY2025-26, cumulative signings past 1,035, and a Finance Act 2026 that consolidated safe harbour categories and streamlined APA administration. APAs and safe harbours are, in effect, taxpayers pre-negotiating their way out of exactly the compressed, algorithm-driven scrutiny process that produced the insurance company's evidence crunch. That's a rational choice if you can afford the APA application fee and multi-year process. It's not a choice available to a mid-sized captive or a regulated entity whose compliance bandwidth is already stretched across sectoral requirements, as this case shows. The open question for practitioners to sit with: as India's tax administration becomes more algorithmic, should the same accountability and explainability standards the government is asking of private-sector AI deployers - logged reasoning, contestability, human-in-the-loop review before adverse action - apply symmetrically to the department's own risk-scoring systems, especially once (not if) that scoring logic extends into TP-specific audit selection and comparable-set flagging? Right now, that symmetry doesn't exist, and the insurance company's three-day notice is what the gap looks like in practice.

Tax Deadlines Need Grid-Style Planning

Monday is deadline day for the roughly two crore taxpayers filing ITR-3 and ITR-4 for assessment year 2026-27, the freelancers, small traders and professionals who got the extra month that salaried filers did not. And true to form, the run-up has produced the same story administrators watch unfold every year: login failures, slow challan updates, and a Karnataka taxpayers' association writing to the CBDT about credentials that work on the third attempt but not the first. The department's answer, reported this weekend, was to keep the helpdesk open through the night rather than move the date.

the Income Tax Department has decided to keep its helpdesk operational 24x7 until 23:59 on August 31st

Read that sentence carefully and it tells you something about how large public systems learn, or fail to. The department already knows, to the day, when the surge will hit; it built the staggered calendar precisely to spread that load, moving business filers to August so July would carry only salaried returns, as the report in BusinessToday on this deadline makes clear. That is real design thinking, and it worked, mostly. But the response to the predictable residual crunch is still human overtime: longer helpline hours, staff on call, a circular reminding everyone to save drafts. What is missing is the habit of treating the last three days of a filing window as a known demand spike to be engineered for in advance, the way a power utility plans for peak load on the hottest afternoon of the year, and not as an emergency to be staffed through after it begins.

I have watched this pattern from inside a tax administration for years: the software teams are excellent, the front line staff heroic on deadline week, and yet the institutional memory resets every season. Nobody owns the question of what tested, reserved capacity the portal needs on day minus one. A public organisation that logs this exact data every single year, hourly login volumes, exact failure points, has no real excuse for surprise. The fix is not more goodwill from the helpdesk. It is a standing peak load protocol, reviewed and stress tested months before the date, treated with the same seriousness a grid operator gives a heatwave.

#IncomeTax #ITRFiling #DigitalGovernance #TaxAdministration #PublicSector #CBDT #GovTech

A Payroll Wedge Just Vanished

India's Comprehensive Economic and Trade Agreement with the United Kingdom enters into force today, and most of the coverage will lead with the 99 percent duty-free access on tariff lines. The more interesting instrument arrived beside it. The Double Contribution Convention exempts Indian professionals on temporary UK assignments (five years, extended from three) from paying into Britain's National Insurance while they continue to contribute at home. The report in India Briefing pegs the annual saving at roughly USD 500 million, covering 90 to 95 percent of Indian professionals sent through Indian employers. That number matters less than the frame. India's largest single export is not steel or leather; it is people-hours priced in foreign currency. Tariff talk fixates on goods, but the real friction in services trade has always been a payroll wedge: mandatory contributions the visiting worker never gets to draw down. Neutralising that wedge inside a trade treaty is a subtler innovation than any tariff schedule, and probably a more durable one. The next FTA worth watching is the one that repeats this move.

#CETA #IndiaUKTrade #DCC #ServicesTrade #FreeTradeAgreement #GlobalMobility #SocialSecurity

The Line Above Four

India's June CPI print landed at 4.38%. It sounds unremarkable. It is not. That is the first time in seventeen months the number has crossed the RBI's four percent target, as recorded in a Bloomberg report on Monday. The reading stayed inside the tolerance band. But a line was quietly crossed, and lines matter in monetary policy the way statute matters in tax administration: not because breaking them is catastrophic, but because they change the standard of proof.

The RBI cut 125 basis points last year and pushed liquidity into the banking system on a scale that would have been unusual in a tighter era. The private capital cycle was supposed to receive that liquidity and put it to work. Then food happened. Nearly forty percent of the consumer basket in India is food, and this year's monsoon has run below normal. The government's own Monthly Economic Review argues the economy is now less exposed to rainfall than before, and that is broadly true in the aggregate. But the price index does not measure the aggregate. It measures households in their kitchens.

From inside a national tax administration, one learns that macro forecasts and taxpayer reality diverge more than models suggest. TDS collections track nominal transactions. Nominal transactions carry inflation inside them. Every basis point of CPI drift shows up months later in the shape of the tax base, in the composition of refunds, and in the arguments taxpayers make about real versus nominal incomes. Inflation is not just a monetary story. With a lag, it becomes a tax administration story too.

The immediate temptation, when a target line is crossed, is to over-read the print. One month is a data point, not a trend. But the WPI figure released the same week printed close to its three-year high. Two indices, drifting apart, telling different stories about the same economy. Wholesale is a producer's world; retail is a consumer's. When they diverge this widely, the middle layer of small firms, informal wage earners and first-year borrowers absorbs the friction.

The takeaway is not for the RBI. They will do what the numbers oblige them to do. The takeaway is for anyone building fiscal, tax or investment plans for the second half of this year. Assume a firmer floor under prices, and design for it now, quietly, before it becomes fashionable to do so.

#IndiaEconomy #Inflation #RBI #MonetaryPolicy #CPI #WPI #TaxPolicy #FiscalPolicy

One Screen, Two Statutes

A small notice on the e-Filing portal this week is worth pausing on. The department has quietly rolled out an integrated payment module that lets taxpayers pay dues under the Income-tax Act, 1961 for periods up to FY 2025-26 and under the Income-tax Act, 2025 for Tax Year 2026-27 onwards, all from a single interface. The notice on the Income Tax Department portal puts it plainly:

Seamless payments now enabled across both the Income-tax Act, 1961 and the Income-tax Act, 2025.

That word, seamless, does a lot of work. India is running two direct tax statutes in parallel, one for closing out old years and one for opening new ones. Every legal transition of this scale creates a temptation to make the citizen learn the transition too, to force them to pick which Act their payment belongs to, to route them through different portals, different challan formats, different mental models. The interesting bit of engineering here is the opposite instinct: absorb the complexity inside the system so that a person paying a demand from AY 2022-23 and a person paying advance tax for TY 2026-27 use the same three clicks. The citizen does not need to know which statute is doing the arithmetic.

There is a wider lesson for public administration in this. The measure of a good transition is how quickly it becomes invisible to the person on the other side of the counter. Officers see two Acts, two sets of rules, two saving clauses, two mental frameworks running simultaneously. The taxpayer should ideally see one screen. That gap, between the complexity we carry internally and the simplicity we present externally, is where administrative craft lives. It is worth building for. And it is worth defending against every impulse to expose the plumbing.

#IncomeTax #IncomeTaxAct2025 #CBDT #TaxAdministration #DigitalIndia #eFiling #TaxReform