Artificial Intelligence (AI) is gradually changing the tax inspection model of the United States government. Not only is the Internal Revenue Service (IRS) using AI to improve audit efficiency, but state governments are also accelerating the adoption of AI analysis tools due to financial pressure. Accountant Ye Junlin introduced on July 18 at the “AI New Wave” forum held by the Los Angeles Taiwan-American Chamber of Commerce that in the future, AI will not replace tax inspectors but rather determine “who tax inspectors should inspect first.” The AI audits of state governments may be faster and more influential than those of the federal government.
Many people mistakenly believe that AI tax inspection means the government will monitor all taxpayers comprehensively, but this is not the case. Currently, the main function of AI is to quickly organize, analyze, and score risks from massive tax, financial, and asset data, screening out cases worth further examination before being handed over to employees to decide whether to conduct an audit.
Ye Junlin emphasized that AI does not increase the number of tax inspections but rather makes tax inspections “more accurate.” In the past, due to limited manpower, many anomalies in reporting may not have received attention. Today, AI can quickly conduct large-scale data comparisons, making it easier to detect cases that were previously overlooked.
The judgment of AI is not based on a single standard but rather establishes a risk model through cross-referencing multiple factors, mainly covering four aspects:
– Comparison with industry averages (such as whether the deduction rate is abnormally high)
– Changes in reported income over the years (such as significant discrepancies between income and past years)
– Cost of living in the area
– The reasonableness of income and expenses (such as significant disparities between reported income and living expenses, or purchases of assets that do not align)
Ye Junlin stated that industries with a high volume of cash transactions, such as restaurants, construction, real estate, consulting services, may likely become the focal points of AI’s attention in the future. However, being flagged by AI does not necessarily mean being audited; it only signifies that the case will prioritize manual review processes.
With the enhancement of AI analytical capabilities, the issuance of CP2000 “Suggested Changes Notice” by the IRS may also increase. Ye Junlin mentioned that CP2000 is not an audit notice but a request for taxpayers to explain or pay the difference when the IRS finds inconsistencies between the taxpayer’s reported information and third-party data like 1099s or K-1s. Previously, due to limited efficiency in manual comparisons, some discrepancies could not be identified promptly; now, AI can complete millions of data comparisons within seconds, making similar notices more prevalent in the future.
Regarding the widespread concern of the Chinese community about reporting overseas assets, Ye Junlin stated that with the promotion of the Foreign Account Tax Compliance Act (FATCA) and the Common Reporting Standard (CRS), the exchange of overseas financial information has become standard practice. The real issue is no longer whether the IRS obtains the data but how AI integrates and analyzes this information.
Ye Junlin cautioned that overseas bank accounts, stock investments, company shares, funds, trusts, and cryptocurrencies may all fall within the scope of AI analysis. If a taxpayer reports low income but shows significant fund flows or holds multiple overseas assets, AI is more likely to discover discrepancies between the reported information and the actual financial situation.
Apart from the federal IRS, Ye Junlin believes that the rapidly emerging AI tax inspection systems of state governments are worth noting. He mentioned that state governments generally face higher financial pressure than the federal government, so their audit motivations are stronger, and they have more direct access to resident data. In the future, state tax audit actions may be more proactive than those of the IRS.
Taking California as an example, in recent years, the state government has been discussing wealth taxes, wealth taxes, increasing sales taxes in certain areas, and strengthening Medi-Cal eligibility reviews. All these indicate that the state government is expanding its sources of revenue through various means, with AI being an important tool to enhance audit efficiency.
Ye Junlin analyzed that state government AI is currently mainly applied in four major areas:
– Resident identification: Matching driver’s licenses, real estate, children’s school records, mobile phone locations, etc., to determine if taxpayers still actually reside in California.
– Sales tax: Deep analysis of e-commerce transaction data.
– Property tax: Comparing actual real estate transaction prices with reported values.
– Payroll tax: Auditing whether companies improperly classify employees as independent contractors to evade taxes.
Ye Junlin estimated future audit trends. Between 2026 and 2028, state government AI audit cases will significantly increase, and the California Franchise Tax Board (FTB) and Employment Development Department (EDD) may enhance the use of AI enforcement. By 2028 to 2030, as the IRS manpower gradually supplements, federal and state government audits will concurrently upgrade, forming a more comprehensive inter-agency data sharing and audit mechanism.
Ye Junlin reminded that the IRS may currently not be able to handle all AI-flagged cases immediately due to limited manpower, but the system will continue to accumulate high-risk cases, and they will be audited in detail once manpower is restored. Since the general case review period is usually three to six years, with potential fraud cases having no time limit, the public should not hold the mentality of “if not audited now, then it’s all clear.”
He concluded that in the age of AI, compliance has become the best risk management strategy, stating, “AI will not replace tax inspectors, but AI will decide which tax inspectors to prioritize. “He advised taxpayers to review their tax information, overseas assets, and related reporting obligations early, ensuring the consistency of all information, which is far better than rectifying issues only after receiving an audit notice, reducing time, costs, and legal risks. ◇