Original article can be found here (source): artificial intelligence
Current business models are becoming heavily reliant on data, analytics and artificial intelligence (AI) to make real-time business decisions. In a time dictated by black swan events and uncertainty, the ability of technology to better understand a complex world and to take decisions that consistently mitigate risks while assuring greater returns will be tested to its limit.
With “social distancing” as the new “corona” mantra, the ability to act and operate remotely through the use of machines will make businesses more resilient. It is also expected that with the flattening of the corona curve, new digital business models will develop, as more and more CEOs fight to keep their businesses alive.
AI Investing and Fund Management is the Way Forward
The fund management industry has been a pioneer in using AI for improving performance and has already been pivoting towards machine-managed funds as opposed to relying on the instincts of its human managers. It is believed that traditional investing is riddled with human bias, whereas machine investing is more objective. Consider the following statistics in relation to the U.S. market:
- Funds run by computers that follow rules set by humans account for 35% of America’s stock market, 60% of institutional equity assets and 60% of trading activity.
- In 2016, quant funds became the largest source of institutional trading volume in the American stock market. They accounted for 36% of institutional volume in 2019, which is up from just 18% in 2010, according to the Tabb Group, a research firm.
- Only 10% of institutional trading is through traditional equity fund managers according to JPMorgan Chase.
- The total value of American public equities is $31 trillion, as measured by the Russell 3000 Index. Computer-managed funds including index funds, exchange-traded funds (ETFs) and quant funds run around 35% while human managers, such as traditional hedge funds and other mutual funds, manage just 24%.
- Of the $18 trillion to $19 trillion of managed assets accounted for, most are looked after by machines. Index funds manage half of that pot, around $9 trillion. According to Bernstein, a research firm, other quantitative equity managers look after another 10–15%, roughly $2 trillion. The remaining 35-40%, worth $7 to $8 trillion, is overseen by humans.
- According to Deutsche Bank, 90% of equity-futures trades and 80% of cash-equity trades are executed by algorithms without any human input. Equity-derivative markets are also dominated by electronic execution according to the Tabb Group.
In contrast, 30–50% of the trades are made through algorithmic trading in India.
The corona push may result in more investment decisions, including trading and holding, being based on statistical correlations and big data crunching. Apart from operating within parameters set by humans for making investment decisions, AI programs are also writing their own investing rules.
Software programs using AI devise their own strategies without human guidance after processing and finding patterns in large amounts of data, and many times such strategies are not completely understood by the human makers either. Such programs which deliver results, without the programmers or funds managers understanding how such a program makes decisions, are sometimes referred to as “black boxes” due to the inability to understand how the software has internalized huge stores of data.
While some quant funds use algorithms to perform data analysis and then call on humans to select trades, many quant funds, however, are pushing automation even further, by using machine learning and AI to enable machines to pick which stocks to buy and sell. In many cases, co-location of servers close to the servers of the respective stock exchanges increases speed of transaction confirmations, which is key to high frequency trading or algo-trading business models.
This is made possible due to increasing processor capability and greater flows of information. An almost infinite supply of new data and processing power is creating novel ways to assess investments. For example, some funds try to use satellites to track retailers’ car parks, and scrape inflation data from e-commerce sites. Eventually they could have fresher information about firms than even the firm’s board of directors and therefore make better investment decisions.
Indian Tax Risks—Taxing Data
But what is the connection between AI-driven investment decisions and Indian tax?
The story started a few years ago when in the wake of digitization of business models, G-20 countries asserted that historical principles of international taxation and allocation of income to countries such as India were insufficient.
Under the current rules, countries like India are not able to tax income of companies operating highly digitalized business models as such companies typically do not have physical presence in India.
Over the last few years, however, the discussion has evolved with the Organization for Economic Co-operation and Development (OECD) and India has been fast(est) to react, with several changes being inserted in the Indian Income Tax Act, 1961. The most recent of these changes now proposes to target income derived (directly or indirectly) from the analysis of data collected in India and seeks to increasingly tax it in India.
The 2020 Budget proposes that in the case of a nonresident who has a business connection (or taxable presence) in India, income from certain activities including sale of services using data collected from a person residing in India or from a person who uses an internet protocol address located in India would be attributable to India and therefore taxable under domestic law.
It is therefore possible that income derived from or through the analysis of any data collected from a person in India could potentially result in tax exposure, for example geo-location data collected from mobile devices or user data collected from applications on mobile devices.
It may also not matter that the data was bought from a third party or that is available in the public domain as long as the data originated in India.
However, it does not appear, under the proposed rule, that all data collected or originated from India should create a tax risk as non-personal data, meaning data that is not collected from or connected to a particular individual, could arguably be outside the ambit of the above provisions (e.g. data “scraped” from the internet, satellite imagery of relevant locations for public companies such as mines, retail parking lots or shipping lanes and other publicly available website data).
However, it is possible that if any such data is directly or indirectly connected with persons in India, use of such data could still potentially create an India tax risk, for instance, it may be possible to argue that retail parking slots relate to persons or individuals, as opposed to satellite imagery of forests.
In the context of many fund managers, especially those managing funds with a clear AI-driven investment or holding strategy, decision making is at the core of services provided by fund managers to their investors.
This leads to a question as to whether provision of fund management services utilizing data collected from India could come within the ambit of the new rules. Also, the use of AI software housed in servers in India that make investment decisions could result in further tax exposures through the possible creation of a server permanent establishment under current law.
AI decision making in India on investment parameters could also raise the prospect of tax risks arising in the form of Place of Effective Management (POEM), according to which a foreign entity is taxed as a tax resident in India on its business income at 40%, if the key commercial decisions are taken in India. In this context, it is important to note that the law does not specify that key commercial decisions should necessarily by taken by a person or an individual and in the context of funds, arguably whether to invest or not could be considered a key commercial decision.
While many funds may have their AI located in servers outside India, the global push towards taxing income in a country where the foreign company is not a tax resident, without the creation of any physical presence, could make the location of the server redundant and therefore create exposure for all fund managers who utilize data collected from India for their decision making.
In such a situation, it is pertinent to note that the Central Board for Direct Taxes has in a paper in 2019 suggested that at least 10–20% of the total income should be attributable to data collected from a person resident in India, depending on the level of participation or user contribution by such person in the generation of the data. The impact of Indian taxes on such an allocation is likely to be high and significantly hinder the economics of fund operations.
Should Fund Managers be Worried?
Nevertheless, such income is taxable in India only if the foreign fund manager has a taxable presence both under Indian tax laws and any applicable double tax avoidance treaty. Currently, under Indian law, foreign fund managers may create a taxable presence in India, without any physical presence, if they conduct transactions in goods, services or property above a certain monetary threshold (these thresholds are yet to be notified); or there is a systematic and continuous solicitation of business from India from a prescribed number of users through digital means; or a systematic and continuous engagement with a prescribed number of users through digital means (the specified value would be on an aggregate basis and based on the value of a single transaction).
While it may be a stretch to say that collection of data from India would result in such taxable presence being created, they could possibly be triggered through the sourcing of deals or making investments in India and making decisions using AI/software located in Indian servers which pose a real tax risk.
Further, it is important to note that if the fund is based in a treaty jurisdiction, under current treaty law it is possible to take benefit of treaty provisions which usually prevent India from levying taxes unless the foreign fund has some form of physical presence in India. However, that benefit would not be available to funds based out of non-treaty jurisdictions such as the Cayman Islands.
Even the benefit of treaty protections available under current law cannot be taken for granted post April 1, 2020, since the implementation of the multilateral instrument in India (a global treaty signed by over 100 countries) will make such treaty provisions contingent on the satisfaction of several tests including the “principal purpose test” (PPT).
According to the PPT, the treaty benefit can be denied if one of the purposes for investing through a jurisdiction is obtaining a tax benefit. The burden is then on the taxpayer to prove that the fund is set up in a particular jurisdiction for other non- tax reasons as well, or risk losing the treaty benefit.
In any case, decision making through AI-based software maintained in servers in India (assuming fund managers have control over the whole or part of the server) is likely to pass the permanent establishment test contained in most tax treaties. Further, as mentioned above, the current trajectory of the global discussions on taxation of digital businesses is likely to result in the dilution of current treaty protections, thereby allowing taxation of such income without the requirement of physical presence in India in the foreseeable future.
Not every crisis is the same; different factors arise that need to be considered while making investment decisions. While the financial crisis in 2008 required a different response, the damage caused by the coronavirus is likely to be offset in the long run overall and the economy will recover.
Funds and fund managers need to be careful taking any investment decisions in the event they are under lock down in India due to the coronavirus, as that may create tax issues by making the fund a tax resident in India. It may also create permanent establishment for the fund in India in some circumstances. As many actions are likely to be reactionary in such unprecedented times, any deviation from standard protocol should be reviewed and any resultant tax exposures should be mitigated at the earliest.
As regards the future, in an increasingly complex and uncertain world where shocks could emanate from weaknesses in the financial systems, pandemics or climate change, the ability to take into account more information and make better decisions is key to being agile.
Such real time decision making is going to be indispensable and also heavily reliant on AI. As tax policy attempts to adapt and keep pace, both at a global and India level, the onus is on the funds to be aware of possible tax risks and mitigate them where possible.
Parul Jain is Co-Head International Tax and Meyyappan Nagappan is a Leader at Nishith Desai Associates, India.
All statistics, facts and figures in this article are taken from publicly available information and sources.
This column does not necessarily reflect the opinion of The Bureau of National Affairs, Inc. or its owners.