National Tax Service's AI Tax Investigations Go Into Full Swing…"Not After-the-Fact Explanation, but Advance Documentation Is Key"

This past July, the National Tax Service (NTS) announced an 'AI Grand Transformation' roadmap to apply AI across the entire national tax administration, from tax consultations to tax investigations and delinquency management.

Starting with building the core infrastructure for this in the second half of this year, the plan is to launch the main project next year and expand the service to all areas by 2028. The core of this change lies not simply in introducing new technology.

Tax investigation is being gradually transformed from a method of verifying individual transactions to a method of comprehensively analyzing vast data—such as fund flows, consumption patterns, transactions among family members, and overseas asset movements accumulated over several years—to detect anomalies.

It amounts to a fundamental change in the paradigm of national tax administration for the first time in about 40 years, since the selection of investigation targets through computerized systems began in 1984.

In the past, investigation targets were selected according to criteria set by people, centered on filing data or ledgers. A representative example is the PCI (income-expenditure analysis) system, which compares the increase in assets and the scale of consumption against declared income to detect imbalances in the asset-formation process.

The NTS's 'AI Grand Transformation' roadmap goes one step further from here. A text mining technique, in which AI itself learns patterns related to evasion suspicions from vast unstructured data such as tax investigation results, adjudication precedents, and filing data accumulated over decades to predict risk levels, will be combined in stages.

If the roadmap proceeds as planned, it is expected that imbalances in the asset-formation process will be identifiable far more precisely than now.

Alongside this, important analytical clues are the reporting data of financial institutions. High-value cash transaction (CTR) reports automatically filed with the Korea Financial Intelligence Unit (FIU) upon cash deposits or withdrawals of 10 million won or more per day, and suspicious transaction (STR) reports filed by financial institution employees, are already major subjects of NTS analysis, and are expected to be even more tightly combined with the AI analysis system going forward.

This data will play a key role in screening cash flows abnormally large relative to declared income, or repeated split withdrawals.

However, as AI tax administration expands, exaggerated information such as 'the NTS monitors all of an individual's financial transactions in real time' is also spreading. The reality is different.

The AI Grand Transformation does not mean building a system that constantly monitors all accounts, but is closer to advancing a tool that analyzes the possibility of evasion based on data collected within the scope set by law to select investigation targets.

Nor does a mere transfer of living expenses among family members or a one-time remittance immediately trigger a tax investigation. It is a structure in which the need for investigation rises only when multiple data points combine to form anomalies—such as an increase in assets that does not match declared income, repeated fund movements, or transactions whose economic substance is unclear.

There is one more point to note here. The fact that AI detects an anomaly does not mean the result immediately signifies tax evasion. AI is merely a tool to screen investigation targets; what ultimately determines whether tax is levied is law and evidence.

Whether a parent temporarily lent funds to a child, whether a family pooled funds to jointly purchase real estate, and whether a transaction between a corporation and its representative was based on a legitimate contract are judged not by AI but by objective evidence submitted during the investigation—such as contracts, loan certificates, interest payment records, and accounting materials.

In the end, even for the same transaction, how clearly one can prove its legal nature determines the outcome of the investigation.

This change is also transforming the paradigm of tax risk management. In the past, after-the-fact response—preparing explanatory materials after a tax investigation began—was common.

In contrast, now that the AI analysis system is being advanced in stages, advance management—systematically leaving behind evidence to support legal legitimacy and economic substance from the stage the transaction takes place—has become essential.

Therefore, in monetary transactions among family members, one must check the drafting of a loan certificate, payment of appropriate interest, and whether withholding tax applies, and in capital transactions of a family corporation, one must closely review appropriate valuation and the contract structure.

Even when acquiring high-value assets, a procedure to confirm in advance whether the declared income and the source of funds are logically connected is necessary.

The tax-saving strategy of the AI era is no longer simply a technique for reducing taxes. Building in advance a structure that can explain the economic substance and legal legitimacy of a transaction with objective materials becomes the most powerful risk management.

To respond to the advancing AI analysis system, a sophisticated legal and tax strategy is required that comprehensively considers not only simple tax knowledge but also review of the legal effect of transactions, the investigation stage, and appeal procedures.

Josebo / Certified Tax Accountant Jung Kyung-ok, Daeryun Law Firm

[Read the full article]
National Tax Service's AI Tax Investigations Go Into Full Swing…"Not After-the-Fact Explanation, but Advance Documentation Is Key" (Go to link)