HMRC uses artificial intelligence to reduce the tax gap

There has been wide media coverage of the French tax authorities using artificial intelligence (AI) to find over 20,000 undeclared private swimming pools. This has resulted in additional tax revenues of approximately €10million as swimming pools can result in higher French property taxes. 

The software developed by Google and Capgemini identified swimming pools from aerial pictures and cross-referenced these to land registry databases. It is anticipated that the total tax take from this project will be €40million once it has completed its national roll out. 

The use of AI by the French tax authorities to support its compliance activities is not unique. HMRC announced in 2018 that it was examining how it could use AI for its compliance activities and complex tax cases.

In 2020, HMRC confirmed that it uses AI to support a number of its activities including:

• identifying risks on some of its large-scale transactional services such as VAT and self-assessment repayments;
• data analytics to identify tax risks and build case packages on taxpayers for HMRC investigators; and
• assimilating large amounts of data for its compliance activities which is used alongside other tools such as geo-mapping.

HMRC shared its use of AI in a paper published by the International Public Sector Fraud Forum (IPSFF) on the use of artificial intelligence to combat public sector fraud. The IPSFF comprises representatives of government organisations from Australia, Canada, New Zealand, United Kingdom and the United States. The purpose of the forum is to share best practice in combatting public sector fraud. The paper confirmed that the use of AI has become an essential tool in detecting and preventing public sector fraud.

HMRC has been using data analytics as a strategic tool for a number of years to tackle tax evasion through its Connect software. Connect was developed with the help of BAE systems and purportedly cost £100 million to develop but has reportedly recovered at least £3bn in taxes. 

HMRC does not share a lot of information about Connect publicly, but in a letter responding to a parliamentary committee enquiring into reducing the tax gap by using third party data, HMRC confirmed that the software is a data matching and risking tool that allows HMRC to cross match one billion HMRC and third-party data items. This allows HMRC to target its compliance activities on previously unidentified relationships between people and organisations and inconsistencies between taxpayers’ lifestyles and their known sources of income.

HMRC does not disclose all of its data sources but Connect has access to the information HMRC holds from tax returns and over 30 third party databases including:

• bank accounts and pensions;
• credit reference agencies;
• foreign tax jurisdictions and information reported under the common reporting standard which covers over 100 foreign jurisdictions;
• DVLA records;
• sales websites such as Amazon and eBay;
• government agencies such as the Border Agency and the Land Registry;
• flight sales and passenger information;
• property websites;
• online social networking accounts;
• credit and debit card accounts;
• online payment accounts;
• Google ‘Street View’; and
• insurance companies.

HMRC is expected to continue to improve its use of AI, not just by learning from its own experiences but also by understanding how other tax authorities and government organisations around the world use AI. As the UK economy inevitably moves to one where even more transactions are undertaken and recorded digitally and electronically, HMRC’s access to third party data will increase. HMRC’s increasing use of AI may lead to more targeted enquiries to prevent and detect tax evasion and reduce the tax gap.