AI Regulation in E-Commerce: Recommendations and Dynamic Pricing

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

Artificial intelligence algorithms already set the prices you pay when you shop online, and regulating this market is one of the greatest legal challenges of our time.

The Rise of Artificial Intelligence in E-Commerce

In recent years, we have witnessed a profound transformation in how e-commerce operates in Brazil and around the world. Online sales platforms have begun using artificial intelligence systems to personalize the consumer experience, from product recommendations to real-time price setting. This technological shift has brought undeniable benefits to market efficiency, but it has also raised complex legal questions that still lack specific regulation.

When we look at how recommendation systems work, we see that they rely on enormous volumes of data collected from users: browsing history, past purchases, geographic location, the device used, the time of access, and even scrolling behavior on the page. This data feeds machine learning algorithms that build detailed profiles of each consumer, anticipating their preferences and directing offers in a highly personalized way. The problem arises when personalization goes beyond convenience and becomes a potentially abusive practice.

In Brazil, the Marco Civil da Internet (Brazil’s Internet Civil Rights Framework) (Law 12.965/2014) and the General Data Protection Law, or LGPD (Law 13.709/2018), establish general principles of transparency, purpose, and consent in the processing of personal data. However, neither of these laws specifically addresses the use of AI in recommendation systems and dynamic pricing in e-commerce. This regulatory gap creates legal uncertainty for businesses and consumers alike.

Dynamic Pricing: How It Works and Why It Raises Concern

Dynamic pricing consists of automatically adjusting the prices of products and services based on variables such as supply and demand, consumer profile, competition, and the timing of the purchase. The technique is not new: airlines and hotel chains have used it for decades. What has changed is the sophistication of the algorithms. Now powered by artificial intelligence, they can adjust prices in a fraction of a second and at the level of the individual consumer.

The main legal concern with dynamic pricing is the potential for price discrimination based on a consumer’s personal characteristics. When an algorithm determines that a particular user is willing to pay more (because they use a more expensive device, for example, or live in a wealthier area), the system can automatically show that person higher prices. When this practice rests on opaque criteria and happens without the consumer’s knowledge, it may violate the principle of objective good faith set out in the Consumer Protection Code (Law 8.078/1990).

Article 6 of the CDC (Brazil’s Consumer Protection Code) establishes as a basic consumer right adequate and clear information about different products and services, with correct specification of quantity, characteristics, composition, quality, applicable taxes, and price. When the price shown to a consumer is the result of algorithmic processing that weighs personal variables without transparency, we have to ask whether this fundamental right is being respected. In addition, Article 39 of the same statute prohibits practices that impose a manifestly excessive advantage on the consumer, which may occur in cases of abusive dynamic pricing.

Algorithmic transparency is not merely a technical issue. It is a legal imperative for ensuring that e-commerce operates in line with consumers’ fundamental rights.

In practical terms, we see that large marketplaces already make thousands of price changes a day across their catalogs. Individual consumers cannot perceive these variations, since they generally have no way of comparing the price they are offered with the one shown to another user at the same moment. This information asymmetry is one of the reasons regulation is necessary.

The AI Regulation Bill and Its Impact on E-Commerce

The Brazilian legislative debate on regulating artificial intelligence has advanced in recent years, with several proposals moving through the National Congress. The text that has progressed furthest sets out principles, rights, and duties for the development and use of AI systems in Brazil, including provisions relevant to e-commerce. Among the central points being followed are the classification of AI systems by risk level, the requirement of algorithmic impact assessments, and the creation of human oversight mechanisms.

The implications for the e-commerce sector are significant. Recommendation and dynamic pricing systems that use sensitive personal data, or that may produce unlawful discrimination, are likely to be classified as high-risk, which imposes additional obligations on their operators. These obligations include detailed technical documentation, algorithmic bias testing, a guaranteed right to an explanation of automated decisions, and mechanisms that allow consumers to contest those decisions.

We also look at the influence of the European Artificial Intelligence Act (AI Act), approved by the European Union, which serves as a reference for Brazilian legislation. The European model takes a risk-based approach that sorts AI applications into categories ranging from minimal risk to unacceptable risk. Under that model, pricing systems that use subliminal or exploitative techniques to manipulate consumer behavior may be treated as prohibited practices.

At the national level, the National Consumer Secretariat (Senacon) and the Administrative Council for Economic Defense (CADE) have already shown interest in monitoring algorithmic pricing practices. The concern is not limited to protecting individual consumers. It also extends to competition: pricing algorithms that monitor and automatically react to competitors’ prices could, in theory, facilitate forms of tacit price coordination, which raises significant antitrust questions.

Consumer Rights in the Face of Recommendation Algorithms

When we examine recommendation systems from the standpoint of consumer law, we find several points of tension that deserve regulatory attention. The first concerns transparency: consumers have the right to know they are interacting with an automated system and what basic criteria drive the recommendations they receive. Article 20 of the LGPD already gives data subjects the right to request review of decisions made solely through automated processing, including decisions meant to define their personal, professional, consumer, and credit profiles.

In practice, though, applying this provision to e-commerce presents considerable challenges. Most consumers are not even aware that the prices and products they see are personalized by algorithms. Even when they are, exercising the right to review runs into the technical complexity of the AI models involved, which often operate as “black boxes” whose decisions are hard to explain even for their own developers.

Another aspect we consider relevant is consent. The LGPD requires that the processing of personal data have an appropriate legal basis, and consent is one of them. Yet the terms of use and privacy policies of e-commerce platforms are often written in generic language, without clearly stating that the data will be used to personalize prices. We believe that consent obtained under these conditions can be challenged as invalid, especially given the legal requirement that consent be free, informed, and unambiguous.

Beyond the individual level, we note that recommendation systems can create so-called commercial “filter bubbles,” in which consumers are repeatedly exposed to the same types of products and price ranges, limiting their ability to make truly free choices. This algorithmic curation of consumption raises questions about autonomy of will (autonomia da vontade), one of the pillars of Brazilian contract law.

Paths Toward Balanced Regulation

Given the current landscape, we believe that regulating AI in e-commerce should seek a balance between effectively protecting consumers and preserving the sector’s capacity to innovate. Overly restrictive regulatory proposals may stifle technological development and hurt the competitiveness of Brazilian companies, while the absence of regulation allows potentially abusive practices to persist.

Among the measures we consider a priority is mandatory algorithmic transparency in language consumers can understand. Platforms should state, clearly and visibly, when the prices displayed are personalized and which categories of data are used for that personalization. This transparency does not require disclosing the algorithm’s source code or trade secrets. It requires communicating the general criteria that influence how the price is formed.

We also recommend implementing “opt-out” mechanisms that let consumers choose to see non-personalized prices. This option, already discussed in European legislative proposals, gives consumers the power to choose whether to be subject to personalization, without eliminating the benefits it can offer to those who prefer to keep it active.

Creating a regulatory body with specific technical expertise to oversee AI systems is another point that deserves attention. This body could be a new entity or a responsibility assigned to existing agencies such as the ANPD (National Data Protection Authority). It should be able to conduct algorithmic audits, evaluate complaints of price discrimination, and impose sanctions proportionate to the violations found.

Finally, we consider investment in consumer digital education essential. Awareness programs about how recommendation algorithms and dynamic pricing work can empower citizens to exercise their rights more effectively. To us, the combination of appropriate regulation, technical oversight, and consumer education looks like the most promising path for artificial intelligence in e-commerce to deliver its potential to create value without compromising fundamental rights.

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

This topic is directly related to AI regulation in, a subject 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.