Regulation of AI-Driven Algorithmic Trading

This page is a translation of the original Portuguese text, provided for convenience. It deals with Brazilian law. If there is any discrepancy, the Portuguese version prevails. Read the original in Portuguese

The rise of artificial intelligence in financial markets calls for robust regulatory frameworks for algorithmic trading, protecting investors and the integrity of the system.

Algorithmic Trading Today in Brazil and Around the World

In recent years, global financial markets have been transformed by the growing use of artificial intelligence algorithms to buy and sell assets. Algorithmic trading, also known as algo trading, relies on sophisticated computer models to execute orders in fractions of a second, analyzing massive volumes of data and spotting patterns that conventional human analysis would never detect. This technological revolution has brought undeniable benefits, such as greater market liquidity, lower operating costs, and more efficient price formation. It has also introduced systemic risks that challenge regulators worldwide.

In Brazil, B3 (Brasil, Bolsa, Balcão, the Brazilian stock exchange) already reports a significant share of algorithmic activity in total trading volume. High-frequency trading (HFT), statistical arbitrage, and automated market making make up a relevant part of the trading ecosystem. The Comissão de Valores Mobiliários (CVM), Brazil’s securities regulator and the main authority over the Brazilian capital markets, has followed these developments with growing attention. It recognizes the need to update the regulatory framework to address the specific challenges that artificial intelligence poses for financial operations.

Looking at the international landscape, jurisdictions such as the European Union, the United States, and the United Kingdom have already made considerable progress in regulating these activities. Europe’s MiFID II, for example, sets detailed requirements for firms that engage in algorithmic trading, including obligations for system testing, risk controls, and operational transparency. This international experience is an important reference for the development of Brazilian regulation, although adjustments to the particular features of the Brazilian market are indispensable.

Legal and Systemic Risks of Trading with Artificial Intelligence

Using artificial intelligence in algorithmic trading creates a series of risks that go well beyond technology and enter legal territory, with significant consequences. The first and perhaps most debated risk is market manipulation. Algorithms can be programmed (or, in the case of reinforcement learning models, can learn) to carry out strategies that amount to unlawful practices, such as spoofing (placing and rapidly canceling orders to create a false impression of demand) and layering (creating multiple tiers of artificial orders). Identifying these practices when they are carried out by autonomous systems poses considerable evidentiary challenges.

Another significant risk is the so-called “flash crash,” an event in which interconnected algorithms amplify market movements out of control, causing abrupt drops or spikes in asset prices within extremely short periods. The most emblematic episode took place in May 2010 on the New York exchange, when the Dow Jones index plunged nearly a thousand points in a few minutes before partially recovering. Similar events, though smaller in scale, continue to occur in various markets, exposing the fragility of existing containment mechanisms.

From a civil liability standpoint, we face complex legal questions. When an AI algorithm causes losses to investors or to the market as a whole, who is responsible? The developer of the algorithm? The financial institution that uses it? The operator who configured it? The chain of liability becomes especially tangled when we consider that machine learning systems can develop behaviors their creators never anticipated, making autonomous decisions that escape direct human control. This opacity in decision-making, often described as the “black box” problem, is one of the greatest challenges for contemporary law.

Regulating AI-driven algorithmic trading requires balancing the encouragement of technological innovation against the effective protection of investors and the preservation of the integrity of financial markets.

Concerns about information asymmetry also arise. Market participants with more sophisticated algorithms and superior technological infrastructure (such as co-location of servers next to the exchange’s systems) gain competitive advantages that can undermine fairness in trading. This disparity raises legitimate questions about the principle of equal treatment among investors and about the acceptable limits of technological advantage in the capital markets.

Brazil’s Regulatory Framework and Applicable Rules

In the Brazilian legal system, the regulation of algorithmic trading rests on several legal instruments that, although not originally designed to address artificial intelligence, provide a basis for governing these activities. Law 6.385/1976, which governs the securities market and creates the CVM, gives the agency broad regulatory and oversight powers that extend to transactions carried out by electronic and automated means.

The CVM has issued instructions and resolutions that touch on the subject. The rules on direct market access (DMA) set requirements for intermediaries that allow their clients to send orders directly to trading systems. These rules include pre-trade controls, limits on quantity and price, and automatic halt mechanisms (circuit breakers) that act as safeguards against anomalous algorithmic behavior.

Law 13.709/2018 (the General Data Protection Law, or LGPD) also has a relevant connection to algorithmic trading when algorithms process investors’ personal data to make automated decisions. Article 20 of the LGPD gives data subjects the right to request review of decisions made solely on the basis of automated data processing. This principle may be invoked where AI algorithms directly affect investors’ financial interests.

The Legal Framework for Artificial Intelligence (Marco Legal da Inteligência Artificial), currently under debate in Brazil’s National Congress, also deserves attention. It could establish general principles for the use of AI in every sector, including the financial market. Among the principles under discussion are algorithmic transparency, the explainability of automated decisions, human oversight, and liability for harm caused by AI systems. Passage of this legislation would be a significant step toward a clearer and more predictable regulatory environment for algorithmic trading.

Self-Regulation and the Role of B3

Beyond the state regulation exercised by the CVM, B3 plays a fundamental role in the self-regulation of algorithmic trading. The Brazilian exchange maintains its own monitoring and control mechanisms, including market surveillance systems that themselves use artificial intelligence to detect suspicious trading patterns. B3’s trading rules set specific requirements for participants that use algorithmic strategies, including registration obligations, compliance testing, and operating limits.

BSM Supervisão de Mercados (B3’s self-regulatory arm) conducts investigations and administrative proceedings concerning potentially improper practices carried out by algorithms. This work faces the constant challenge of distinguishing legitimate (even if aggressive) algorithmic behavior from genuinely manipulative practices, a task that demands ever more sophisticated technical expertise.

Regulatory Trends and Future Outlook

Looking at global regulatory trends, we see converging movements that will likely shape the development of Brazilian rules in the coming years. The European Union, through the AI Act (Artificial Intelligence Regulation), has classified AI systems used in critical infrastructure (including financial markets) as “high risk,” subjecting them to strict requirements for conformity assessment, technical documentation, and human oversight. Regulators in many jurisdictions have come to regard this risk-based approach as a benchmark.

In the United States, the Securities and Exchange Commission (SEC) and the Commodity Futures Trading Commission (CFTC) have stepped up their oversight of the use of AI in financial markets. Recent proposals include requiring intermediaries that use predictive data analytics technologies (including AI) to eliminate or neutralize conflicts of interest that could harm investors. This approach reflects a growing concern that algorithms may be used to optimize financial institutions’ revenues at the expense of their clients’ interests.

In Brazil, we expect regulatory change to follow a few main lines. The first is stronger algorithmic governance requirements, with more detailed obligations for documenting, testing, and continuously monitoring the AI systems used in trading. The second involves broader transparency duties, possibly including an obligation to flag orders generated by algorithms and to disclose information about the algorithmic strategies used (in aggregate form, protecting trade secrets). The third concerns better accountability mechanisms, with a clearer definition of the chain of responsibility when AI algorithms cause harm to the market or to individual investors.

Compliance and Governance for Financial Institutions

For financial institutions that use, or plan to use, algorithmic trading with artificial intelligence, building robust compliance and governance programs is becoming increasingly essential. These institutions are advised to adopt clear internal policies on the development, testing, and deployment of trading algorithms, with precisely defined roles and responsibilities. Keeping detailed records (audit trails) of algorithmic decisions makes it easier both to supervise internally and to respond to any regulatory requests.

Stress testing and simulations of adverse scenarios are indispensable for assessing how algorithms behave under extreme market conditions. Likewise, layered risk controls (pre-trade, during execution, and post-trade) help mitigate the risk of unwanted events. Ongoing training for the teams involved, combining knowledge of technology, finance, and regulation, rounds out the governance framework needed to operate in an increasingly complex environment.

Ethical Implications and the Balance Between Innovation and Protection

Beyond strictly legal questions, regulating AI-driven algorithmic trading raises fundamental ethical questions about the role of technology in financial markets and in society. We have to ask whether it is desirable for capital allocation decisions (which affect companies, jobs, and entire economies) to be gradually handed over to autonomous systems whose inner workings are not always fully understood by their operators. The concentration of technological capacity in a few market participants raises legitimate concerns about democratizing access and about distributive justice.

At the same time, we recognize that technological innovation in the financial market has the potential to produce significant social benefits, such as lower intermediation costs, more efficient allocation of resources, and wider access to investment. The regulatory challenge lies precisely in finding the balance between fostering innovation and protecting fundamental values such as systemic stability, fairness among participants, and confidence in the way markets work.

With that in mind, we advocate a regulatory approach that is technology-neutral (regulating outcomes and conduct, not specific technologies), proportionate to the risks identified, adaptive (able to evolve as fast as the technology), and collaborative (involving regulators, market participants, academics, and civil society in drafting the rules). Only with this broad and balanced view can we build a regulatory framework that meets the challenges artificial intelligence poses to the financial market.

This subject is directly related to artificial intelligence and regulation, a topic we cover in a dedicated article.

This subject is directly related to AI regulation in the, a topic we cover in a dedicated article.

As informações deste artigo são de caráter informativo e não substituem consulta jurídica individualizada. Cada caso possui particularidades que exigem análise específica.