AI in Tax: Automation, Risk and Transparency

Tax in the Deal 2026

The relationship between taxpayer and revenue authority is undergoing rapid, fundamental transformation. Tax compliance has always been retrospective: the taxpayer collects data, completes computations and submits returns months later; the authority reviews those submissions and decides whether to enquire. That is changing.

Tax compliance is being re-engineered to operate alongside new systems of commerce characterised by 'live' transactions and the centralisation of data and processes. Tax has ceased to be a discrete, back-office, after-the-event reporting function; it is a continuous, increasingly automated presence in commercial operations and critical to business decision-making.

This change has been underway for years, driven by digitalisation and businesses searching for greater efficiency and commercial advantage. Tax authorities have had to respond, both to secure their own efficiencies and to obtain the data they need to assess compliance and protect national revenue. The introduction of artificial intelligence (AI) will see the pace of change accelerate.

For those involved in M&A activity, the shift raises practical questions including:

How robust is a target's digital tax infrastructure?
At what point does a deficiency in a target's tax technology become a material issue affecting valuation?
How does a buyer obtain comfort on the accuracy and sufficiency of a target's digitalised tax data?
How will a newly acquired business rank under a tax authority's algorithmic risk-assessment system?
Does a transaction result in an increased risk of challenge or dispute with a tax authority?
How should warranty and indemnity provisions reflect these emerging risks?
How will the role of AI in tax affect the tax insurance market?

According to the OECD, more than 70% of advanced tax administrations are deploying AI in enforcement. HMRC is investing heavily to be a leader in adopting and scaling AI. A company's operations, arrangements and structures are increasingly visible to authorities worldwide. It used to be the case that HMRC operated in 'silos'; they found it difficult to align administration across different departments dealing with different parts of the tax system. That has been changing in recent years, but AI allows tax authorities to interrogate the vast amount of data available it collects, flagging weakness and inconsistencies.

A company's tax risk profile will increasingly be shaped by the quality of its data and its systems' ability to gather, process and report them. In addition, the expanding information-gathering powers of HMRC, allied with AI, make genuinely real-time reporting (RTI) a realistic prospect.

Data Data Everywhere

Revenue authorities have spent decades building data collection and information exchange capabilities. Enhanced transparency has long been seen as a powerful early warning system to help combat evasion and abusive avoidance, both domestically and internationally.

The UK has been at the forefront of these efforts, introducing regimes such as the Disclosure of Tax Avoidance Schemes (DOTAS), the Senior Accounting Officer (SAO) and Uncertain Tax Treatment (UTT) obligations. DOTAS formed the blueprint for DAC 6 and the OECD's Mandatory Disclosure Rules. The UK is now going further by looking at improving the quality of third-party data it can collect and the way it is used in tax administration, and in late 2025 introduced a 'Strengthened Reward Scheme' to incentivise whistleblowers to report serious avoidance activities.

The UK has also been an early adopter of international reporting standards such as the Common Reporting Standard (CRS) and Country-by-Country Reporting (CbCR). CbCR, developed as Action 13 of the OECD/G20 Base Erosion and Profit Shifting (BEPS) project, gives tax authorities visibility of a multinational's global activities, profits and taxes paid, and now forms the basis for Pillar Two compliance. It has evolved into a public reporting obligation in the EU, Australia and other jurisdictions (although not yet the UK). Transparency was one of the three guiding principles (alongside coherence and substance) of the BEPS project — designed as a weapon against avoidance, it has also generated the vast, standardised datasets that now power AI-driven enforcement.

The evolution, and layering, of reporting regimes is demanding more, and more granular, data. The demands are also becoming more frequent and increasingly require more standardised formats. This is exactly the type of data that AI relies upon.

In 2024, the OECD's Inventory of Tax Technology Initiatives¹ estimated that of the tax authorities of the 54 members of the Forum on Tax Administration:

Around 80% obtained data directly from taxpayer business systems, some sent automatically machine-to-machine without human involvement; and
More than 70% were using AI to enhance effectiveness and efficiency, with the most common use cases being fraud detection and risk assessment.

Taxpayer businesses will also adopt AI to drive their own efficiencies. However, the depth, breadth and organised nature of data that is flowing to tax authorities globally, and their ability to train AI engines on that data, should not be underestimated.

Meeting the compliance obligations that result will inevitably require businesses to adopt more sophisticated technologies to gather, process and deliver tax data. While businesses can leverage the same data for their own analytical needs, maintaining control and human oversight of the process will be vital. At the same time, the window to assess the accuracy of reports before submitting them is being squeezed. In a deal context, the same tension applies: more tax data ought to be available during diligence but its accuracy is harder to assess without understanding the efficacy of the technologies that produce it.

The same transparency that was designed to combat avoidance also creates reputational risk for compliant taxpayers. The volume of information available to authorities and the public, combined with AI's ability to interrogate it, heightens reputational, valuation and execution risks. To illustrate with a simple example, a low effective tax rate in a jurisdiction, taken out of context, can be misconstrued as aggressive avoidance, triggering public scrutiny and authority enquiries, irrespective of commercial justifications. This has clear consequences for M&A activity, valuation, risk allocation and disputes.

Delivering on its Transformation Roadmap² (first published in July 2025) "to becoming a world-class tax, customs and valuations authority", HMRC is rapidly scaling its AI capabilities. The Roadmap sets out a strategy to prevent non-compliance, promote compliance and target enforcement activity on areas of perceived risk. Wider and faster adoption of technology and AI are central, with HMRC recognising that modernising how it collects, processes and uses data is critical.

HMRC's July 2026 Progress Update³ details how AI is being woven into the UK tax system. HMRC's Annual Report and Accounts 2025-26⁴ claims that AI and advanced analytics enabled the protection and recovery of £10bn in tax between 2025 and 2026 — the original expectation was that such revenue would only be delivered by 2029–30. That represents £1 in every £9 of the £90.4bn increase in tax revenues, and £1 in every £5 of the approximately £50bn of 'compliance yield' collected for the first time.

By March 2026, HMRC had issued over 28,000 Microsoft Copilot licences with plans to reach 50,000 during 2026; 47,000 employees completed AI learning modules in 2025-26. Copilot is being used to summarise customer calls, triage complaints, draft caseworker briefings and support compliance activity — a rollout HMRC describe as one of the largest in UK government. HMRC has also launched PATH (Pioneering Advanced Technology in HMRC), applying a "start small, scale fast" approach to deploying technologies, and two accelerator initiatives designed to "unite policy, operations and digital expertise" in priority areas.

In May 2026, to accelerate its efforts, HMRC awarded a contract worth £175m over 10 years to Quantexa to modernise its data foundation, analytical capacity and AI capabilities. Under the agreement, HMRC hope to obtain a clearer, connected view of the data it already holds and collects, to improve performance (by making quicker and better-informed decisions), identify tax at risk, and strengthen controls. It is all focused on facilitating wider transformation effort to help close the 'tax gap' and improve customer experience. Quantexa described the contract as "a blueprint for how the UK government deploys AI at scale".
Alongside this technology, the government is recruiting 5,500 additional HMRC compliance officers by 2029-30 (over 3,000 are already in post). In a hint at what might be one of their 'priority areas', HMRC has also recruited over 20 external experts and redeployed 200 staff to tackle offshore non-compliance by wealthy individuals and plans to publish a strategy for 'Further Closing the Wealthy Tax Gap' (in Autumn 2026).
HMRC reports that 78% of customer interactions are now digital and are targeting 90% by 2030. It holds 97 million unique records in a new Central Customer Registry, providing a holistic view of the customer. The government has announced the introduction of 'e-invoicing' from April 2029, plans to modernise Corporation Tax systems by 2028, and is standardising company tax returns to support the future administration of corporation tax. All these actions help HMRC gather more data, more quickly, in standardised formats that empower its AI capacities.

The government is also exploring where digitalisation allows reporting windows to narrow and payment dates to accelerate. A good example is the new Securities Transfer Tax (replacing stamp duty and SDRT, planned for 2027), built on a digital foundation with the potential for shorter reporting and payment windows than is currently the case for stamp duty.

Corporate taxpayers should expect corporation tax compliance to change materially. Reporting obligations will grow while payment windows shrink. The ability to control processes, gather data effectively, assess risks, and report and pay taxes in near-real-time will be critical components of efficient tax functions.

As always, there are opportunities and risks.

The increasing demand for data means taxpayers must organise and report information using appropriate systems. The same data collection is an opportunity because it should enable businesses to run AI-assisted analysis prior to submission — for example, to stress-test transfer pricing arrangements, cross-border structuring assumptions, and Pillar Two exposures against the risk-profiling parameters used by revenue authorities. In a notable 2025 decision, the First-tier Tribunal in the UK concluded that taxpayers are entitled to transparency on when and how AI is used by HMRC in its decision-making.

The risks lie in assuring data quality. Taxpayers remain under a duty to take reasonable care when filing; failure can result in tax-geared penalties. The growing importance of tax technology, and speed of digitalised compliance obligations, does not change that duty. Exercising reasonable care will include implementing strong governance controls, including human oversight, over tax processes. AI tools are only as powerful, and their outputs only as good, as the quality of the data that trains them — poor collection and classification will distort accuracy and feed directly into a tax authority's risk assessment of the business.

In an M&A context, digitalisation and adoption of AI will have a direct impact across the deal cycle:

The sophistication of a target's data capabilities will be a material concern in assessing tax risks. A business that can demonstrate robust control over its data flows and adherence to authority standards is better placed to give buyer assurance and less likely to face future enquiries.
Consideration will need to be given to the coverage of risks presented by digitalisation — including data quality from collection to submission, use of AI in tax, preparedness for new digital regimes, and the target's risk profile with authorities. W&I insurers will scrutinise technology-related tax risks and may seek exclusions or require additional information.
Integration of the target's systems with the buyer's represents heightened risk in an environment of near-real-time reporting. It must be carefully managed to avoid disruptions in data flows at the moment when compliance oversight is weakest. Mechanisms for conduct and oversight of digitalised compliance during any transition will need to be agreed.

The defining characteristic of a resilient tax function will be clean data integrated with rigorous, human-led governance. The OECD's Framework for Trustworthy AI in Tax Administration advocates a 'human-in-the-loop' architecture to preserve accountability — a principle that applies equally to taxpaying businesses.

Transparency has long been a tool against abusive avoidance, but it has also generated vast data for authorities to interrogate and share globally. With AI, authorities can now leverage that data to understand taxpayer risk profiles better than ever. Control of the same data can, however, be used by taxpayers to make their operations more efficient. Mutual transparency in AI usage will be critical.

In M&A, tax due diligence must now extend to digital infrastructure, data integrity and algorithmic governance. A company's exposure to HMRC enforcement is increasingly determined by how it interacts with HMRC's digital tools. The strength of that interaction will impact diligence, risk assessment and pricing. Post-completion integration and rationalisation of data processing will be increasingly important. The businesses that adapt to the data-rich tax landscape fastest will realise genuine commercial advantages.

1 OECD (2025), Tax Administration Digitalisation and Digital Transformation Initiatives, OECD Publishing, Paris, https://doi.org/10.1787/c076d776-en

2 HMRC (2025), Corporate report: HMRC's Transformation Roadmap, HM Revenue & Customs, HMRC's Transformation Roadmap - GOV.UK

3 HMRC (2026), Corporate report: HMRC Transformation Roadmap – Progress Update 2026, HM Revenue & Customs, HMRC Transformation Roadmap — Progress Update 2026 - GOV.UK

4 HMRC (2026), Corporate Report: HMRC annual report and accounts: 2025 to 2026, HM Revenue & Customs, HMRC annual report and accounts: 2025 to 2026 - GOV.UK

5 Thomas Elsbury v The Information Commissioner [2025] UKFTT 915 (GRC)