The law is an organism that adapts to economic reality. It did so when it regulated the criminal liability of legal entities, in order to modernize the regulation of the liability of increasingly complex and opaque corporations.
In short, it is likely that it will have to do so again if the companies governed by artificial intelligence agents and without human executives proliferate.
When fiction becomes a bill
Argentina has not waited for the debate to mature in academic forums. The Government has sent to Congress a draft of a new General Law of Companies that recognizes, for the first time, what is known as Autonomous Enterprise, a figure that has until now been nonexistent in law, but which encompasses two figures:
On one side, the "Automated Company," a company with full legal personality, capacity to contract, and its own patrimonial responsibility, which can operate completely autonomously through algorithms, without the need for employees for its ordinary functioning, nor human control or supervision.
And, on the other side, the "Decentralized Autonomous Operating Company," originating from the crypto ecosystem: organizations that are totally or partially autonomous and decentralized, which set their own rules of governance, decision-making, and profit distribution through smart contracts deployed on blockchain, with participation represented in tokens.
If it prospers, Argentina would become one of the first jurisdictions in the world to give legal status to the company without humans.
What Roman law could not foresee
The liability of legal entities has already forced the law to develop specific attribution mechanisms, based on compliance models and a specific concept of due diligence.
The transformation of this branch of law was considered by the more traditional doctrine as a considerable exercise of legal engineering, because even a traditional public limited company ultimately has a board of directors or executives, identifiable natural persons behind each relevant decision.
However, the criminal liability of the legal entity has been consolidating thanks to relevant jurisprudential pronouncements. Now, the existence of a self-managed company, governed by AI agents, poses a qualitative leap on that aspect, and not a few doubts.
For example, who responds when the business decision is made by an algorithm, trained and deployed by a diffuse community of token holders spread around the world, without there being an administrative body in the classical sense?
The regulatory fronts that are opening
In the corporate and commercial sphere, recognizing legal personality for an entity without human administrators forces a rethink of some traditional concepts such as, for example, who holds representation before third parties when the ordinary operation is decided by a language model? How is the duty of diligence and loyalty articulated when there is no flesh-and-blood administrator to demand it from? What kind of responsibility can be demanded from an algorithm?
In terms of taxation, the Tax Administration needs a domicile, an identifiable representative. A distributed organization in code, without a physical headquarters or effective address in the classical sense, raises doubts regarding central concepts of traditional taxation, and opens the door to much more sophisticated structures than those we already know.
There should also be reflection on the eventual criminal responsibility of the entity. In this sense, if corporate criminal responsibility already required imagining the organizational fault of a company, the imputation to a company governed by AI requires delimiting responsibilities in one or several of the people involved at some point in the decision chain or, simply, we need a new legal figure that responds to these new needs.
An opportunity to not be late
Spain and the European Union have been regulating artificial intelligence for years from the risk perspective —the AI Regulation is good proof of this— but they have barely begun to address the corporate question: what happens when the company itself, and not just its tool, is autonomous.
The Argentine initiative, with all the caution that a project still in parliamentary debate deserves, brings some of these doubts to the surface. In particular, regarding aspects ranging from money laundering to the evasion of corporate controls. In any case, it has the merit of putting the question on the table before anyone else in the Spanish-speaking world.
The history of criminal law of legal entities teaches us something useful: the legal fictions that yesterday seemed to be incorporated into the legal system when economic reality makes them inevitable.
Currently, just as yesterday happened with the criminal liability of a legal entity, a company governed by AI agents without human executives may seem today a rarity or a technological experiment. However, it is not.
The Spanish legislator would do well to start studying the problem before the question reaches him already turned into litigation. Or, worse yet, that these types of entities begin to develop and operate de facto, based in jurisdictions with a low level of supervision, in the absence of a clear response from the European legal system.
about the author:
Francisco Pérez Bes is deputy of the Spanish Agency for Data Protection. In addition, he was a partner in the Digital Law area of Ecix Group and is a former Secretary General of the National Cybersecurity Institute (INCIBE).