In the landscape of modern tax administration, the principle that the burden of proof rests on the taxpayer is a fundamental tenet. This principle dictates that it is the taxpayer’s responsibility to substantiate their tax positions, despite the significant disparity in available resources between individual taxpayers and the government.
While this framework aims to ensure tax compliance and integrity, the reality of implementing this principle can be far more challenging, especially if the South African Revenue Service (SARS) relies on potentially flawed third-party information and increasingly sophisticated technologies such as artificial intelligence (AI).
The Challenges of Third-Party Information
It is possible for significant issues to arise for taxpayers with SARS depending on third-party information for tax assessments. Third parties include employers, financial institutions, and other reporting entities such as medical aids and other reporting entities, who are responsible for providing data on taxpayer’s income, deductions, and other financial activities. This system is designed to streamline tax reporting and enhance accuracy, but it does not always come without its challenges.
Third-party information may be prone to inaccuracies and errors. For instance, an employer might incorrectly report an employee’s income due to clerical or administrative errors, a bank could report the interest earned on a savings account incorrectly. When SARS relies on incorrect third-party data to determine a taxpayer’s tax liability, proving the tax liability to be incorrect can get more challenging than one would hope. This is especially the case where the taxpayer lacks direct access to or control over the third-party records.
The Impact of Artificial Intelligence
SARS has adopted the use of artificial intelligence (“AI”) to handle and analyse vast volumes of data. AI and machine learning are helping SARS identify and address criminal activities and instances of non-compliance.[1]
While AI systems are able to deliver speed and enhanced efficiency, there are still risks at hand, such as over-surveillance, job disparity, and lack of transparency and accountability[2]. AI systems are only as reliable as the data they process and the algorithms that are used. If the input data is incorrect or the algorithms are flawed, the AI can produce inaccurate assessments resulting in taxpayers finding themselves contending with erroneous AI-generated conclusions.
The burden of proof vs SARS policy
In cases where third-party data is the cause of the incorrect assessment, our experience, as dispute resolution experts, is that SARS will require the third party to confirm the data is incorrect or even correct the data before they are willing to accept that the assessment is incorrect.
To some extent, this is understandable. Allow us to explain by way of a rudimentary example:
Assume a taxpayer worked outside the country, but the employer, on the taxpayer’s IRP5, uses a source code reflecting the income earned to be in South Africa. Clearly, the IRP5 casts doubt on any contention by the taxpayer that the services were, in fact, rendered outside SA.
Enter, however, the standard of proof.
In tax disputes, the standard is “on a balance of probabilities”. This simply means that the taxpayer carries the burden of proving that, on balance, something is probably true. Let’s then return to the example above.
If the taxpayer provides sufficient evidence that indicates he was working outside the country, then on balance, he did, regardless of what is stated on the IRP5. SARS, however, will, in our experience, often insist that the IRP5 must be changed or confirmation must be given from the employer that the IRP5 is incorrect. The reason for this, it seems, is that SARS appears to hold taxpayers to a higher standard of proof where disputes between taxpayers and SARS are in the pre-litigation phase.
Here, in the -prelitigation phase of a tax dispute, SARS policy seems to be that the standard of proof is closer to: ‘beyond reasonable doubt’. Indeed, in the example above, if the employer does not fix the IRP5 there may still be reasonable doubt about whether the taxpayer worked outside the country, despite evidence from the taxpayer to the contrary.
Now if the taxpayer can meet this higher standard of proof because they can get the third-party data corrected then do that. It is often the path of least resentence.
But what if the taxpayer can’t get the third-party data sorted out? What if the taxpayer has no idea how the AI tool came the conclusion it did? Does it mean the taxpayer has no case and should simply accept the incorrect assessment?
Absolutely not! In our experience, however, it does mean that those taxpayers’ dispute resolution journey will be hard and long and will often require the assistance of tax dispute resolution experts. A tax dispute resolution expert knows the difference between SARS policy and the law (and contrary to popular belief, the two are not always aligned) and when to push and when not to push a dispute ahead.
These cases tend to eventually find their way into the litigation phase of tax dispute resolution where the issues are purely those of facts and law, and not of SARS’ and policies and procedures and where the actual standard of proof is, in our view, applied correctly.
Conclusion
The increased use of third-party reporting and AI tools, whilst probably great for improving efficiency in the tax system on a large scale, may have the opposite effect on some taxpayers. Granted, these cases will probably be the exception, but it is exactly because it will be the exception that the struggle will probably be harder for those taxpayers. Policies and procedures are not written for the exception but for the norm, and where, in any large organisation, there is an occurrence of something outside the norm, it’s harder to fix. Even more so where that large organisation is in the public sector and happens to hold the public purse.
[1] ITWeb. (2024, February 26). SARS doubles down on AI, machine learning after revenue gains. ITWeb. Retrieved from https://www.itweb.co.za/article/sars-doubles-down-on-ai-machine-learning-after-revenue-gains/rxP3jqBEyeoMA2ye
[2] Accounting Weekly. (2024, February 27). SARS expands detection capabilities with advanced technology. Retrieved from https://www.accountingweekly.com/trending-news/ail3048frmizb9lmggy42difk7i5u8#:~:text=In%20an%20impressive%20demonstration%20of%20technology’s%20power%20in
