The capital markets are also majorly subject to transformations due to AI – even if this is not visible at first “with the naked eye”, the analysis of complex data and the automation of processes is also happening here, with human intervention being less and less required. Both the labor market and intellectual property data (where applications for AI patents have grown exponentially in recent years) signal this with sufficient certainty (1).
Whether we are talking about optimizing investment decision processes or discovering new opportunities for capital market growth, the influence of AI is being felt, and the efficiency of financial markets can be dramatically enhanced in this way.
Although what AI brings to financial markets (increased productivity, a greater degree of finesse in the design of investment frameworks and portfolios, improved forecasts of investment returns, more accurate quantification of risks) is, in principle, auspicious, optimism should still be tempered by at least the same level of caution.
The most relevant example for the need for caution in the use of algorithms is the flash crash since 2010, when within about 10 minutesprices crashed and the Dow Jones fell about 9%, only to bounce back minutes later. The destabilization of the market occurred, apparently, due to the use of some algorithms and the parameters they were set on, the cause of the problem not being fully understood even to this day.
The 2010 incident is a textbook example of black-box tradingthat is, that type of trading based on algorithms whose mode of operation is unknown. These are automated trading systems that operate without human intervention and process enormous amounts of buy/sell orders at millisecond speeds. The most interesting aspect of these trading systems, however, comes from the fact that the way of operation is generally secret, the logic behind which the respective algorithm works is not revealed to the user client (and sometimes cannot be explained even by the creator of the algorithm).
Why do these matters matter, and not only to those concerned specifically with capital markets?
Such trading systems have not only been present in capital markets for some time, but currently dominate them in terms of trading volume compared to classical trading mediated by human decisions.
Or, in such conditions, to look at black-box trading only as a problem that belongs exclusively to hedge funds or investment companies would be imprudent, since the effects of such systems spread further than that. For example, pension funds or asset managers (which are more “tangible” entities for the individual customer, the common man), are institutions that trade in the same markets as the systems mentioned above, competing, directly or indirectly, with the performance of black-box and high-frequency trading (HFT) algorithms, or perhaps even using them.
Practice already proves that market manipulation through algorithmic strategies, especially through HFT, works successfully (this fact, by the way, also raises disputes regarding the liability of those participants who use such methods). (2)
But, traditionally, the notion of “market manipulation” was, in turn, based on the notion of “intent”: a market actor decides to distort prices, deliberately.
But when the problem arises of executing transactions by means of algorithms, which not only have capabilities incomparably superior to human ones (in terms of trading volume, speed, lack of sentiment to influence a trading decision), but it also works in a black-box trading system (i.e. totally non-transparent), how else can we relate, from a legal point of view, to the notion of “intention”? Implicitly, how can we verify the existence of market manipulation, with the consequence of the possibility of later holding accountable the persons responsible for the losses thus caused?
The algorithm, depending on the degree of independence under whose empire it was designed, executes orders or even makes autonomous decisions. Thus, with a “classical” program, one can infer the intent of the algorithm's creator, depending on the content of its code, but with a black-box AI system, such a check no longer works. It is true that we can know what the overall objective was (eg profit maximization), but how the AI achieves that goal, including whether it does so through market manipulation, may be impossible to understand, both before and after the trade is executed (and that includes by its creator). Therefore, legal tests based on intent become virtually impossible to apply. (3)
Although real progress has recently been made in the way of understanding black-box models, in the sense of researching them as “objects” that are observed from the outside (without addressing the method of direct interaction with them), the applicability of these advances to the capital market is still far from becoming a reality. Since this new understanding is based on observing the pattern from the outside and finding some “triggers” that lead to a certain decision of the respective AI, however, the speed that characterizes the field of capital markets will make it extremely difficult to apply this method in this case as well. Reaction times of the order of milliseconds will not allow the observation of the behavior of the black-box model “from the outside” to be carried out in real time, which will practically lead to a post-factum evaluation of the trading decisions made and to the impossibility of observing, in this case, the “triggers” of the decision.
Even in the absence of the “intention” of the trading algorithm, the effects of an AI trade in this system can be similar to those caused by market manipulation: artificial falling/increasing of prices, abnormal trading volume, market shock (resulting in the birth of extreme, cascading volatility), as already happened in real cases.
In such a situation, the question of liability arises from a legal point of view: will the programmer of the algorithm, the company that uses it, the person who designed the trading strategy answer? Since the very process by which a black-box algorithm arrived at a particular result is completely opaque, determining who bears legal responsibility becomes an elusive goal. The EU AI Act, which appears to be superior in terms of regulatory requirements for 'high-risk' AI system providers compared to MiFID II, although it moves in the right direction and attempts to establish comprehensive AI regulation, will not, however, be able to clearly answer the question “who answers” when an opaque algorithm performs problematic transactions.
Given the extremely high reaction speed and interconnectedness of the markets, whether we are talking about investment funds, listed companies or simply individual investors, it is absolutely necessary that all these actors understand very well and accept who exactly they are sharing the ring with. Although the game already exists and is played with high stakes, the referee is either late or absent altogether. It remains to be seen, under these circumstances, who will be declared the winner.
Footnotes:
(1) International Monetary Fund, 2024, Global Financial Stability Report: Steadying the Course: Uncertainty, Artificial Intelligence, and Financial StabilityWashington DC, October, p. 77
(2) Alessio AZZUTTI, Wolf-Georg RINGE, H. Siegfried STIEHL, “Machine Learning, market manipulation, and collusion on capital markets: why the “black box” matters“, Penn Law: Legal Scholarship Repository, available SSRN: https://ssrn.com/abstract=3788872
(3) Yavar BATHAEE, “The Artificial Intelligence black box and the failure of intent and causation”, Harvard Journal of Law & Technology Volume 31, Number 2, Spring 2018.
An article signed by Ingrid-Amelia Apetrei, Managing Associate, STOICA & ASSOCIATES – [email protected].
Article supported by STOICA & ASSOCIAțII
