KPMG says existing tax rules can handle generative AI, no reboot needed

What KPMG is really saying

KPMG’s core argument is simple enough to travel well in client meetings: generative AI does not require a tax-system reboot. Instead, the questions that matter most can already be handled through the existing stack of international tax concepts, including transfer pricing, nexus, and permanent establishment.

That framing matters because a lot of the splashier debate around AI taxation sounds bigger than the practical challenge. For tax professionals, the work is not to invent a new AI tax species. It is to map AI-driven activity onto rules that already govern cross-border profit allocation, value creation, and business presence.

Myth versus reality

The myth is that AI has broken international tax. The reality is more prosaic, and more operational. Existing rules already force companies and advisers to ask where value is created, what functions are performed, what assets are used, and which entity earns the residual return.

KPMG’s report, titled in part “No need to reboot: GenAI fits the tax stack,” pushes readers toward that practical lens. That is especially relevant inside a Big 4 environment where tax teams are fielding client questions about digital business models while still managing the usual pressures of busy season, promotion-cycle expectations, and the partner-track push for visible technical depth. AI is now part of mainstream international tax planning, not a niche theory exercise that can be left to policy specialists.

Why the old rules still do most of the work

The OECD’s transfer pricing framework remains the anchor here. The OECD says transfer pricing is grounded in the arm’s-length principle and is the global standard for pricing related-party cross-border transactions. That means tax teams still have a familiar framework for asking whether an AI-enabled group structure is allocating profit in line with the functions, assets, and risks actually being carried by each entity.

That matters because generative AI can scale operations without the same headcount or physical footprint firms used to need. A company can serve more customers, automate more workflows, and expand across borders faster than its traditional staffing model would suggest. The tax question is not whether AI is new in an abstract sense. It is whether the current rules can still identify where the economic return belongs when the underlying business has become lighter on labor and heavier on code, data, and intangible assets.

The real pressure points for KPMG tax teams

The pressure points are not theoretical. They sit in the familiar but difficult areas of tax work: intangibles, profit allocation, and cross-border structure.

• If AI is generating value from proprietary data, models, or software, the tax team has to decide which entity owns or exploits those intangibles.
• If an AI platform allows a business to expand without extra people on the ground, the question becomes whether the profit split still matches the local substance.
• If AI-supported services are delivered into a market without a large footprint, the discussion turns to nexus and permanent establishment.
• If tax authorities challenge the arrangement, the audit file has to show why the chosen allocation is defensible under current rules.

That is where KPMG professionals will spend time with clients: not on whether AI should be taxed in some brand-new way, but on whether current structures still stand up when machines do more of the work once done by people.

Why policy debates are heating up anyway

The broader policy conversation is moving for a reason. The OECD says more than 135 jurisdictions joined the October 2021 agreement on the two-pillar solution for international tax reform, and it describes that effort as an update to a tax system that was no longer fit for a globalized and digitalized economy. That background is important because it shows tax policymakers are already working from the assumption that digital business models can strain older frameworks.

At the same time, the United Nations Tax Committee is actively testing the boundaries of the current model. In October 2025, it established a subcommittee on the taxation of the digitalized and globalized economy, and that subcommittee’s mandate explicitly includes whether services related to emerging technologies such as artificial intelligence, machine learning, and the internet of things are adequately addressed by the UN Model Tax Convention.

The UN work also includes looking at services permanent establishments and their relationship to Article 12AA in the UN Model Tax Convention. For practitioners, that is not a side issue. It is a sign that the international debate is shifting from whether AI is real to whether the existing treaty concepts are broad enough to absorb it.

What this means for compliance and audit risk

For KPMG teams, the most immediate impact is in controversy and risk management. The practical test will be whether an AI-driven structure can be explained cleanly under existing doctrines and documented well enough to survive audit scrutiny.

That is especially important because tax administrations are becoming more digital too. The OECD’s “AI in tax administration” work shows governments using AI for internal processes, fraud detection, forecasting, and tailored public services. In other words, the same technology that is changing client operations is also changing how authorities identify anomalies, prioritize cases, and challenge returns.

That should sharpen, not soften, the compliance burden. If a company uses AI to scale across borders with fewer people, the tax file needs to show where the functions sit, how the return is measured, and why the profit allocation makes sense. In practice, that means stronger documentation, more careful transfer pricing analysis, and tighter alignment between business narratives and legal entities.

The takeaway for tax professionals

The biggest mistake would be treating AI taxation as a future problem that can wait for a new regime. The more realistic view is that the work is already here, and it looks familiar: classify the activity, identify the value drivers, test the cross-border footprint, and defend the allocation under existing rules.

For KPMG tax professionals, that makes GenAI less of a policy curiosity and more of a routine advisory issue with high stakes. The firms and clients that will be best prepared are the ones that stop asking whether tax rules need to be reinvented and start asking whether the current framework can still explain how AI creates value, where that value sits, and how much of it belongs in each jurisdiction.