On 25 March 2026 the UN Committee of Experts on International Cooperation in Tax Matters discussed the proposed workstream for the guidance on tax administration and AI. The subcommittee presented its planned work on the guidance for comment and approval.
The Tax Committee is mandated to develop guidance for developing countries on AI use in the tax administration. This would take a modular form, starting with fundamental principles and building on them. The guidance would use examples and case studies. The content of the guide would be practical, dealing with fraud detection, risk assessment, taxpayer services and other functions of the tax administration. The relevant subcommittee would collaborate with subcommittees dealing with related topics such as tax and gender.
The proposed guide would consist of five chapters. The first chapter would cover issues around AI adoption by developing countries and examine the policy rationale. The second chapter would look at considerations for enabling the tax administration to use AI, such as human resources, legal knowledge, use of data, required technology, infrastructure and the governance framework.
The guidance would then examine the potential uses of AI including compliance, audit, detection of abnormal items, and dispute resolution. Consideration would be given to ethical and legal safeguards, looking at risks such as gender bias, errors and system failures. The data produced by the AI must be credible.
The target audience for the guidance would be decision-makers in developing country tax administrations. The guidance would take a modular approach, helping tax administrations at various levels of digital maturity. This approach would have the flexibility to be helpful to a broad range of tax administrations and to maintain its relevance in a fast-changing technological environment.
The subcommittee would ensure that the work is coherent and consistent with work on indirect taxes, tax and gender and wealth tax.
Comments from the participants highlighted problems with connectivity, as illustrated by the recent ITU Facts and Figures 2025 report by the International Telecommunication Union. Although high numbers of people are able to use the internet, around 36% of people worldwide sometimes have difficulty getting connected due to infrastructure problems. This must be taken into account.
Another issue is the training of staff within tax administrations. There are large gaps in technical abilities in digital functions. An OECD study on governing with AI in 2025 noted that 60% of public bodies have staff that are not well trained in new technologies. The subcommittee will therefore need to look at the digital divide and the level of ability to use the new technology within tax administrations.
The business sector contributors pointed out that taxpayers are using AI, and this can create a structural power imbalance. Tax administrations could require taxpayers to be more transparent on their AI use by documenting the AI models they are using and how they are being applied.
The issue of digital sovereignty may become relevant. The largest AI companies are concentrated in one country and other tax administrations are likely to use this technology. They therefore need to protect their sovereignty and ensure that they know that is happening to their data. Another issue that will be important is that AI is very energy intensive. The subcommittee will need to look at who can use AI and how much they will need to pay.
