Tax Authorities Now Audit With AI. The Explainability Standard Is Already Here.
EY's October 2026 Controversy Watch confirms it: tax authorities now use AI, cross-border analytics, and real-time mandates. Disputes start earlier. Documentation gaps that took years to surface now take weeks.
Editorial source review — November 1, 2026. This machine-generated article has been reviewed, but is not certified accurate. Source-backed corrections are incorporated where established. Other legal, survey and product claims may remain unverified; check applicable primary authorities before relying on a deadline or tax position.
Findings in the original version: OECD's September package supports consistency / GIR updates. It does not prescribe the article's three-point / five-point ETR audit trigger thresholds or prove all major authorities use the same AI reviews.
EY's October 2026 Global Tax Policy and Controversy Watch landed with a finding that every F500 tax director needs to act on this week. Tax authorities in every major jurisdiction are now deploying AI, advanced analytics, and cross-border data-sharing infrastructure to review taxpayer data in ways that fundamentally change how disputes arise, how early they escalate, and what standard of documentation survives scrutiny.
The report's core thesis is blunt: disputes are moving earlier in the process. The window between filing and audit initiation is compressing. And the evidentiary standard that authorities are applying is no longer just "did you have documentation?" It is "can you explain, in structured data terms, exactly how you arrived at this position?" That is the explainability standard. It is here. Your current documentation posture was built for the old one.
What EY's October 2026 Report Actually Says
The EY Global Tax Policy and Controversy Watch is published monthly and tracks real developments across tax authorities worldwide. The October 2026 edition focuses specifically on how authorities are using digital tools to reshape the controversy landscape. Three developments stand out.
First, authorities are applying AI to transaction-level data in real time. This is not hypothetical. Australia's ATO, the UK's HMRC, Germany's Bundeszentralamt für Steuern, and the IRS are all running analytics programs that flag anomalies at the transaction level — before an audit is formally opened. When a pattern in your books triggers a flag, the authority already has a hypothesis. You are responding to their data model, not presenting yours from scratch.
Second, cross-border data sharing is operational. The OECD's Automatic Exchange of Information framework has been live for years. What is new is the analytical layer sitting on top of it. When authorities in multiple jurisdictions share transaction data and run it through coordinated models, a transfer pricing position that looked defensible in isolation becomes exposed in aggregate. EY flags this explicitly: companies need to assess how cross-border analytics changes their exposure on historical positions.
Third, the report specifically ties the OECD/G20 Inclusive Framework's September 11, 2026 Global Minimum Tax package to controversy readiness. The package includes updated GloBE Information Return guidance and a framework for the full legislative review process. Every multinational that filed a GIR for 2024 or 2025 should treat that filing as a live document under scrutiny, not an archived submission.
The Transfer Pricing Time Bomb
Transfer pricing is where this shift hits hardest. EY's report says that as authorities expand AI and analytics capabilities, transfer pricing positions built on comparables analysis from 2019 or 2021 are increasingly vulnerable — not because the analysis was wrong then, but because the analytical standard for what "arm's length" means is being recalibrated against real-time transaction data from comparable companies that authorities now have access to.
Think about what that means in practice. Your 2022 transfer pricing study said your intercompany royalty rate was arm's length. The authority in the relevant jurisdiction now has transaction-level data from your 10 closest competitors, aggregated under their analytics platform, and their model says your rate is an outlier at the 85th percentile. You have a documentation package. They have a data model. The dispute is asymmetric from the start.
The EY recommendation is to review how future technologies challenge historical positions. I would frame it more directly: audit your own transfer pricing documentation with the same analytics tools authorities are using. If your positions do not survive that internal review, they will not survive an external one. The question is only when.
Evidence preservation is the second dimension. Authorities are not just checking current-year positions. They are correlating data across multiple years to identify systematic patterns. A single-year documentation package is insufficient if the pattern of your positions over five years tells a story that your annual filings do not. Clarity Tax's approach to this is to treat each year's position as a node in a knowledge graph — connected to prior-year positions, underlying assumptions, and the factual record that supports each. That is the infrastructure that survives a multi-year analytics-driven audit.
What "Explainability" Means for Your Provision Team
The shift to an explainability standard has a specific technical meaning for how you run your ASC 740 provision and your disclosure process. It means the authority expects to be able to trace any tax position — effective tax rate component, deferred tax asset, uncertain tax position — back through a chain of structured data to the underlying transaction and the rule applied to it.
Most corporate tax functions cannot do this today. The provision is built in a spreadsheet. The supporting data sits in an ERP. The rules are in a PDF or in someone's head. When an authority asks "show me how you computed the deferred tax on this intangible transfer," the answer requires assembling data from three systems and producing a narrative explanation that a spreadsheet formula does not generate on its own.
The explainability standard requires a different infrastructure. The tax position has to be built from structured, traceable data — not assembled after the fact. The rule applied to each item has to be recorded at the time of application, not reconstructed from memory during an audit. This is exactly the problem that a tax knowledge graph solves: every concept, every rule, every position is a node with defined relationships to the data that supports it. When an authority asks for the explanation, the graph already contains it.
This is not a nice-to-have architecture decision. It is the difference between surviving an AI-enabled audit and spending 18 months reconstructing your position under adversarial conditions.
The GloBE Information Return as a Controversy Document
The OECD's September 2026 GloBE guidance update is directly relevant here. The GIR is not just a compliance filing. It is a structured data submission that tax authorities will run analytics against. The jurisdictional breakdown of effective tax rates, the constituent entity allocations, the top-up tax calculations — all of it is now machine-readable data in the hands of multiple revenue authorities simultaneously.
EY's report notes that the updated GIR guidance clarifies how authorities should use GIR data. Read that carefully: authorities are receiving formal guidance on how to use your GIR filings as an analytical input. Your GIR is an audit trigger document if your numbers are outliers against authority expectations.
For multinationals with operations in 30+ jurisdictions, the GIR creates a new kind of transfer pricing exposure. If your Pillar Two effective tax rate in a given jurisdiction diverges from the jurisdiction's statutory rate by more than a threshold the authority considers normal, that divergence becomes an audit lead. The documentation standard for justifying that divergence is the explainability standard — not a narrative memo, but traceable structured data.
What to Do Monday Morning
1. Run an analytics review of your own transfer pricing positions. Use the same comparables data that authorities have access to. If your rates fall outside the interquartile range on current market data, treat that as an open controversy item, not a future risk. 2. Audit your GIR filings for jurisdiction-level ETR outliers. Any constituent entity with an effective tax rate more than 3 percentage points below the 15% GloBE minimum — or above the jurisdictional statutory rate by more than 5 points — should have a documented, data-traceable explanation ready before the next audit cycle. 3. Map your documentation infrastructure against the explainability standard. For each material uncertain tax position, can you trace the position back through structured data to the underlying transaction and rule? If the answer requires a spreadsheet and three meetings, you have an infrastructure gap. 4. Update your evidence preservation protocol. Multi-year analytical patterns are how AI-driven audits find systematic issues. Preserve contemporaneous documentation for each year's positions — not just current-year work papers. 5. Brief your board's audit committee. The EY report makes clear that the timeline for disputes has compressed. Positions that historically would have been reviewed in year three of a statute of limitations window are now flagged in the filing year. Audit committee members need to understand that controversy readiness is now a current-period risk management function, not a deferred one.
Clarity Tax Technologies builds the data infrastructure that makes the explainability standard achievable — deterministic tax logic applied to structured data, every position traced to the rule and the transaction that supports it, the knowledge graph that survives an AI-enabled audit. This is what "tax as an operating system" means in practice: not just automation, but auditability by design.