Automation to audit the rich: Can AI bring tax evaders in line?

• Author:
• Author name

  Quantumrun Foresight
• October 25, 2023

Insight summary

Governments worldwide, including China and the U.S., are exploring the use of artificial intelligence (AI) to modernize tax systems. China aims for full automation by 2027, focusing on tax evasion among the rich and social media influencers. In contrast, the U.S. struggles with auditing the wealthy due to reduced IRS budgets and the use of legal loopholes. Salesforce has developed an AI Economist, a tool using reinforcement learning to explore fair tax policies. While promising, the technology raises concerns like increased public surveillance and resistance from wealthy individuals and corporations who may fight automation in taxation.

Automation to audit the rich context

China’s State Taxation Administration vowed to ramp up using AI (2022) to identify tax evaders and give them the harshest punishment under the law. To improve monitoring, China is moving ahead with the development of the Golden Tax IV system, under which company data and information from owners, executives, banks, and other market regulators will be linked and available for tax authorities to investigate. In particular, the country is targeting social media content creators and influencers earning millions of dollars from online streams. China hopes to implement full automation by 2027, using the cloud and big data. China’s wealthy are also anticipating larger tax payments this year (2022-2023), owing to President Xi Jinping’s “common prosperity” campaign.

Meanwhile, taxing the wealthy in the US continues to be an uphill battle. In 2019, the IRS acknowledged that it’s more cost-effective to tax low-wage earners than go after the large corporations and the top 1 percent. The agency declared that since the ultrawealthy have an army of the best lawyers and accountants at their disposal, they are able to take advantage of a variety of legal taxation loopholes, including offshore accounts. The agency’s budget has also been reduced over decades by Congress, leading to suboptimal staffing levels. And while there is bipartisan support to increase the agency’s funding, manual work will not be enough to combat the resources of multimillionaires.

Disruptive impact

Automating tax policies is a complex and often controversial topic. But what if there was a way to make it less political and more data-driven so that it’s fair for everyone? Enter the AI Economist – a tool developed by researchers at technology firm Salesforce that uses reinforcement learning to identify optimal tax policies for a simulated economy. The AI is still relatively simple (it can’t account for all the complexities of the real world), but it is a promising first step toward evaluating policies in a novel way. In one early result, the AI found an approach maximizing productivity and income equality that was 16 percent fairer than a state-of-the-art progressive tax framework studied by academic economists. The improvement over current US policy was even more significant.

Before, neural networks (interconnected data points) were used to manage agents in simulated economies. However, making the policymaker an AI promotes a model in which the workers and policymaker adapt to each other’s behaviors. Because a strategy learned under one tax policy may not work as well under another, reinforcement-learning models had difficulty with this dynamic environment. It also meant that the AIs figured out how to game the system. Some employees learned to cut their productivity to qualify for a lower tax bracket and then increase it again to avoid paying taxes. However, according to Salesforce, this give-and-take between workers and policymakers provides a simulation more realistic than any previously built model, with tax policies typically set and are more often beneficial for the wealthy.

Wider implications of automation auditing the rich

Possible implications of automation being used to audit the rich may include: 

• Increased research on how AI can collate, synthesize, and execute tax filings.
• Countries like China issuing stricter tax regulations on its large corporations and high-earning individuals. However, this may lead to increased public surveillance and intrusive data gathering.
• More available public funding to reinvest in public services of all kinds.
• Increased public institutional trust in government agencies to apply the law and taxation equitably.
• Large corporations and multimillionaires pushing back against automated taxation with increased spending on lobbyists, using data privacy and hacking concerns to counter the technology’s use.
• The wealthy hiring more accountants and lawyers to help them circumnavigate automated taxation.
• Technology firms increasing investments in developing machine learning solutions in the tax sector and partnering with tax agencies.

Questions to comment on

• Do you have experience using automated taxation services?
• How else can AI help in managing tax information and systems?