Can artificial intelligence curb tax evasion?

From bogus bills and fraudulent claims on the input tax credit to loopholes that help evade tax, the Goods and Services Tax System in India has its hands full with grave concerns on tax evasion. While the Government is struggling to meet its GST revenue targets, tax evaders are working in full swing to avoid paying their dues.

Tax officials have detected Rs 37,946 crore worth of tax fraud in 2018-19 and Rs 6,520 crore in the April-June period of 2019-20. FY 2018-19 saw 1,620 cases of fake invoicing involving Rs 11,251.23 crore, leading to the arrest of 154 persons. These figures are expected to grow, which means taxpayers are likely to face high tax rates, ultimately leading to more statistics around tax evasion. The whole process is a vicious cycle which can only be curbed with stringent measures, strict penalties and a robust anti-evasion crew. But experts are of the opinion that even the stringent measures and determined authorities are not going to be enough to nip tax evasion in the bud.

In her Budget 2020 speech, Union Finance Minister, Nirmala Sitharaman quoted, "We wish to open up vistas for a vibrant and dynamic economy with a gentle breeze of new technology." While new technology is a welcome move for efficient filing of tax returns, the finance minister in her poetic words also hinted on use of technology to help deter thousands of crores worth of tax evasion. AI tools can provide improved services like automated tax filing and can also help educate taxpayers to comply with tax filing. The question is, can AI tools help the Government against tax fraud and tax evasion? Let's understand what technological measures could be a part of the GST system in the near future. Let's start with understanding what machine learning is.

Machine Learning to curb tax evasion
In a nutshell, machine learning is the computer's ability to see hidden or emerging patterns in existing data. These patterns can help improve performance without especially programming the system and can be used as an analytical tool by analysts. Imagine using such information to figure out which citizens have been skimming on paying their taxes. Tools that use predictive modelling can enable tax authorities to study fraud cases from history and figure which cases are correlated most with their current cases. It's sort of similar to predicting a player's tell at a poker game - to the untrained eye, the data might mean nothing, but a machine will have a totally different perspective. With the help of AI and robotics, defaulters who are paying less tax or not paying tax at all can be easily found by analyzing and monitoring the financial data of the person.

As a part of India's measures to curb tax evasion, the Government is currently looking at invoice, and input tax credit matching using AI tools and any instances wherein returns have a mismatch of more than 10% are being flagged for identification and inquiry. Recently, the State Government of Delhi commissioned an AI tool developed by two US-based Indian researchers to identify bogus firms made up by traders to evade taxes under Value Added Tax returns from 2017. Similar to this initiative, the Government plans to make use of data analytics and AI tools to red-flag tax evaders and fake refund claims

The use of artificial intelligence and machine learning against tax fraud is already a popular norm among European countries, including Denmark, France and Norway. By using AI tools, Denmark was able to catch around 60 of every 100 cases involving tax fraud - an amazing feat considering Denmark lost close to 2 billion euros in losses in 2018 alone. Norway and France have also made use of AI tools as preventive measures against tax fraud, with France actually using AI tools to comb through social media to detect tax evasion. India is keen on joining the AI bandwagon, and if it does enter the AI anti-evasion scene with technological guns blazing, the economy should hopefully see a drop in tax evasions in the near future.