🎧 What happens legally when it is no longer a human clicking “buy”, but an AI agent independently selecting and purchasing a product or service? In episode #144 of “Alles Legal – Fintech-Recht kompakt”, we explore who is legally responsible for these decisions, why the user’s mandate is crucial and where the boundaries of autonomous AI agents lie. Listen now!
Table of Contents
When AI agents make purchasing decisions
AI agents can do far more than search for information and recommend suitable offers. In agentic commerce, they can make decisions and initiate purchases independently within predefined parameters.
But what does this mean for the existing legal framework? Moderator Dana Wondra from Payment & Banking discusses this question with Frank Müller, Partner at Annerton and author at PayTechLaw.
At the heart of the episode is a fundamental question: to whom should a decision be legally attributed when an AI agent has made it autonomously?
Can an AI enter into a contract?
Imagine you want to fly to Barcelona and give your AI agent only a few parameters: a maximum price of €300, a morning departure and no stopovers. You leave the choice of the actual flight to the agent.
This changes a crucial part of online commerce. Instead of a human ultimately selecting an offer and clicking “buy”, the agent can take this final step.
Under current civil law, however, an AI cannot itself become a contracting party or act as a legal representative. A declaration of intent must still be attributable to a natural or legal person. The question therefore becomes which legal mechanism can be used to attribute an autonomous agent’s actions to its user.
In this episode, Frank Müller discusses several possible approaches, ranging from traditional rules on representation to messenger models, computer-generated declarations and so-called blank declarations.
The mandate defines the agent’s scope
One particularly interesting approach is the mandate model. Here, users do not need to determine every individual decision in advance. Instead, they define a framework within which their AI agent is authorised to act independently.
The scope depends on the specific mandate: How much is the agent allowed to spend? Which merchants can it buy from? What conditions must a product or service satisfy?
The key principle is that an AI agent should not be able to do more than the user has demonstrably delegated to it.
This makes technical solutions for defining and recording mandates, permissions and limits in a machine-readable format increasingly important. If a dispute arises, it must be possible to establish which powers the user actually granted.
Who is liable if an agent exceeds its mandate?
The importance of these limits becomes particularly clear when something goes wrong. What happens, for example, if an agent books a €450 flight despite having been given a €300 spending limit?
Whether the resulting contract can be attributed to the user and who is responsible for exceeding the mandate depends, among other things, on the legal model applied and whether the breach of the mandate was apparent to the parties involved.
For providers of agentic-commerce solutions, this leads to an important conclusion: the objective should not be maximum autonomy, but controllable and demonstrable autonomy.
Users, agents, providers and merchants need to be clearly identifiable. Mandates and limits must be clearly defined, and decisions need to be documented in a way that makes them traceable.
Many of the necessary legal instruments already exist. The greater challenge is adapting them to autonomous AI agents and the new technical processes they introduce.
And concluding the contract is only the beginning. How can an AI agent actually make the subsequent payment? That is the subject of the next episode of “Alles Legal – Fintech-Recht kompakt”.
About this podcast
Alles Legal – Fintech Recht Kompakt delivers sharp, weekly insights into legal and compliance matters in the world of banking. (in German only)
This podcast is a collaboration between Payment & Banking and PayTechLaw.
Each Wednesday, we unpack the legal developments shaping the financial world – clearly, concisely, and without the legal jargon.
Since 2021, PayTechLaw authors and Annerton attorneys have brought depth and clarity to complex topics.
Whether it’s PSD3, DORA, or FiDA – we provide the legal context you need.
In 20 minutes. No detours.
