Consent, responsibility and algorithms: payment rights

The payments sector is at a new inflection point.

Artificial intelligence is about to enter a market that represents 65.7% of payment transactions made with non-monetary instruments: €195 billion every six months. We all know ChatGPT, Claude, or Gemini. Soon we will be able to make purchases directly through these (and other) tools. For example, we could ask it to recommend some headphones and buy the ones it selects directly within the conversation, or even the ones that the AI chooses based on our preferences.

In fact, it is the advent of agentive artificial intelligence that drives this new revolution in the way we shop. Unlike conventional virtual assistants, which respond to questions or generate content, this new generation of AI is capable of making decisions and carrying out specific actions with minimal human intervention.

In the realm of payments, this implies a substantial shift as it opens up the possibility for AI not only to recommend a product but also to complete the transaction on our behalf. Major players in the sector, such as Visa, Mastercard, and American Express, are already working to integrate this technology into their services. The goal is to move towards models in which the purchasing process is gradually automated, reducing human intervention to a minimum: the decision to buy a product. This technological development is part of a broader process in which artificial intelligence is simplifying an increasing number of daily tasks. If users already trust algorithms to obtain information, manage their schedules, or receive personalized recommendations, the next step is to extend that trust to the world of transactions. Delegating the purchasing process to machines raises a critical question from a legal and regulatory perspective. Who assumes responsibility in the event of an error: the machine that made the purchase or the user who made the decision?

There is no simple answer to these questions. The current regulations were not designed with autonomous artificial intelligence agents in mind. The EU Artificial Intelligence Regulation 2024/1689 (the most comprehensive legislation on this matter in the world) does not mention these agents but classifies them in the general category of 'artificial intelligence systems'.

At the same time, the rules governing payment services are currently being updated. The new European directive on this issue (PSD3) and its implementing regulation (PSR) are still in the parliamentary process. And despite this update, they do not explicitly address scenarios involving the complete automation of purchasing decisions.

The current regulatory framework is based on a premise that is beginning to come under pressure, namely, that the person making the purchase is the one making the decision. The advent of agentive AI challenges this assumption and forces a reconsideration of traditional concepts such as consent or responsibility. Tech companies are advancing at a pace that legislators cannot keep up with, leaving consumers in a situation where the applicable rules are not yet clearly defined, but the technology is already prepared for it. However, there is a principle that is likely clear and should serve as a guide. Under no circumstances should AI systems be used to exempt companies from their responsibilities to consumers and users. The progressive automation of purchases is a growing trend.

The challenge will be to ensure that this technological advancement is integrated into a safe and reliable legal framework in the market.

Read the full article here.