When machines converse without us: human exclusion, governance of the invisible and the European regulatory challenge
By Antonio Tejeda Encinas, President of CEA Digital Law (Euro-American Committee of Digital Law) | META Channel Corporation
In sectors such as automated logistics, high-frequency financial trading and the interoperability of digital platforms, machines already talk to each other without human mediation. They exchange data, weigh up options, optimise processes and execute decisions with immediate legal and economic consequences. This phenomenon, which until recently seemed the stuff of science fiction, is now a central part of global technical infrastructure. And yet, in many of its critical aspects, the law has still not addressed it.
At CEA Digital Law, we see this form of autonomous interaction between artificial intelligence systems as a structural shift in the principle of liability. The recently adopted EU Artificial Intelligence Act (AI Act) sets up a pioneering framework to classify, supervise and restrict the use of AI systems according to their level of risk. However, the Act does not resolve – nor does it fully attempt to – the question of effective traceability when decisions are made between interconnected systems without direct human intervention.
What is more, the recent withdrawal of the proposed AI Liability Directive has left a worrying gap. Strict liability, the principle of attribution and compensation mechanisms remain anchored in models designed for human beings or traditional corporate entities, not for autonomous flows of algorithmic decision-making. The law therefore faces a choice: evolve to understand and control these invisible interactions, or be reduced to punishing the consequences without being able to anticipate them.
This mismatch has tangible effects that META Channel Corporation, in its work representing Latin American and global companies in Europe, is identifying ever more clearly. In many cases, decisions affecting the companies it represents – goods being held up, contractual clauses triggered automatically, exclusion from platforms – stem from autonomous interactions between AI systems in different jurisdictions, with no transparency, no explanation and no effective complaints channel. The company, and the people behind it, are left trapped in an automated architecture that decides on their behalf, with no accessible human counterpart to answer for it.
That is why, at CEA Digital Law, we propose moving towards a new legal category: the governance of autonomous interactions, combining:
- Cross-system traceability obligations.
- Mandatory legal audits of autonomous decision-making flows.
- Presumption of shared liability among providers, developers and users of connected autonomous systems.
And at META Channel Corporation, we propose a preventive strategy based on adapting contracts, documenting digital flows and monitoring automated interactions externally, enabling companies to identify legal risks before they materialise. We are facing a new paradigm: the conversation is no longer between humans, nor even between humans and machines, but between machines themselves. And if the law does not adapt to this invisible language, we will lose the ability to govern the systems that govern the world.
