The automated company in the draft of the General Companies Law: challenges, opportunities, and risks of a disruptive figure.

I. Introduction The draft General Corporations Law (the "Reform Bill"), submitted to the National Congress in May 2026, introduces, among its most audacious innovations, the concept of the automated company. Section 14 of the Reform Bill defines it as any company, of any type provided for by law, that "carries out its corporate purpose through autonomous algorithmic systems or artificial intelligence agents, without requiring employees under employment contracts or human resources for its ordinary operation." The only formal requirement imposed by the provision is the statutory record of automation and the inclusion of the term "Automated" in the corporate name. A reading of the provision reveals that, for the first time in the history of Argentine corporate law, the existence of a company that unfolds its economic activity without human beings operating it in an ordinary manner would be enabled; a company whose managerial will no longer stems from the deliberation of directors, managers, or agents, but from the execution of code. This is of particular relevance because Argentine corporate law built its core categories—legal personality, the separation between ownership and management, corporate governance, the liability of directors, the protection of minorities—on the assumption of the human subject. The corporation (sociedad anónima), the preeminent vehicle of large enterprise, is historically explained as an answer to the problem of separation between those who contribute capital and those who manage the business: the board of directors as a professional management body, the shareholders' meeting as the governing body of the owners, the statutory audit committee (sindicatura) as a oversight body, and a delicate system of checks and balances aimed at protecting the investor and third parties against potential abuses by those managing others' assets. That entire conceptual edifice presumes human actors: owners who wish to share in profits, managers who owe loyalty and diligence, minorities demanding information and protection. However, the Bill finds its basis in the fact that, at present, artificial intelligence has ceased to be a mere support tool and has become a factor of structural transformation in business management. Algorithmic systems are no longer limited to processing information or automating repetitive tasks; they can make investment decisions, enter into contracts, set prices, select counterparties, and manage risks with increasing levels of autonomy.

The truth is that, while legal scholars had begun to delineate the duties incumbent upon directors who make use of these tools, the Reform Bill takes a step further and, with perhaps excessive audacity, contemplates the extreme scenario: no longer the human director assisted by AI, but the company that dispenses with human directors in its day-to-day operation. Within this framework, the purpose of this paper (1) is to argue that the automated company, as conceived in the Reform Bill, poses profound tensions with structural principles of corporate law that are not—and cannot be—resolved through mere formal adaptations. This is not, of course, a technophobic objection or a nostalgic defense of the status quo. Corporate automation holds real and valuable opportunities, which we will highlight throughout this discussion: a reduction in opportunistic managerial discretion, greater speed and potential traceability of operations, and the emergence of previously unthinkable business models (2). But these opportunities will only materialize if the concept is built on sound legal foundations. And the central thesis of this article is that the bill has inverted the logical order: it has created the corporate vehicle before resolving the preliminary issues upon which its viability depends, namely: the legal status of the acts of an artificial intelligence agent, the human point of attribution for decisions, and the asset guarantees against damages.

II. The Automated Company in the Reform Bill II.1. Legal Definition and Requirements As anticipated, Section 14 of the Reform Bill defines the automated company by two traits: a functional element and a negative element. The functional element is the performance of the corporate purpose "through autonomous algorithmic systems or artificial intelligence agents." The negative element is operating "without requiring employees under employment contracts or human resources for its ordinary operation." The conjunction of both defines the type: it is not enough for the company to make intensive use of technology; its ordinary operation must be able to dispense with human resources. The formal requirements are sparse. It is required, first, that the declaration of automation be expressly stated in the bylaws; and, second, that the corporate name include the term "Automated." The purpose of this latter requirement is protective: to alert third parties contracting with the company to the particular nature of the entity with which they are engaging. The substantive requirement is the liability regime: "The automated company is liable with its own assets to third parties for damages caused by its autonomous algorithmic systems or artificial intelligence agents." This statement constitutes, as developed below, the core of the problem. As an initial point, it should be noted that the company's asset liability adds nothing to the general regime: every legal entity is liable with its own assets (3). What the concept fails to resolve is what happens when those assets prove insufficient and there are no jointly and severally liable human directors to answer with their own. Now, the Reform Bill exhibits, on this point, an evident tension. Indeed, Section 102 of the Reform Bill provides that "The management body may use artificial intelligence systems or algorithms for the execution of operational functions or decision-making," but clarifies that "Their use neither excludes nor limits the liability of the directors, nor does it constitute an exemption from the duty to configure and supervise the system and its results." It is crystal clear that Section 102 assumes the existence of human directors who use AI as a tool and retain their duties of configuration and supervision intact; it enshrines, in short, the model of hybrid governance. In contrast, Section 14 allows for a company that operates "without human resources for its ordinary operation." How can both rules be reconciled? Who exercises the duty of configuration and supervision under Section 102 in the "pure" automated company of Section 14, if by definition it may lack permanent human directors? II.2. Distinction from DAOs The Reform Bill contemplates, in distinct sections, two concepts that non-specialist opinion tends to conflate: the automated company (Section 14) and the Decentralized Autonomous Operational Company or DAO (Sections 258 et seq.) (4). The automated company operates through algorithms that, however autonomous they may be, respond to a centralized architecture: there is—or should be—a control point from which the system is configured, trained, updated, and, if necessary, shut down. Automation is a mode of executing the corporate purpose, not a form of governing the entity. Capital ownership and ultimate decision-making power remain—at least in theory—in the hands of shareholders or partners, who may be natural or legal persons. A DAO, by contrast, is structured "in a wholly or partially autonomous and decentralized manner, according to the rules of governance, operation, decision-making, and profit distribution provided in its constitutional document or protocol" (Section 258). Its logic is that of smart contracts and distributed governance: decisions are adopted through tokenized voting mechanisms among member owners, without a hierarchical center of power. It is significant that the bill, aware of the risk of total depersonalization, has imposed on the DAO two safeguards that it does not require of the automated company: legal representation entrusted to one or more human beings (Section 260), who act before third parties and bind the company in all acts requiring human intervention; and the requirement that the system guarantee traceability and the identification of member owners in accordance with due diligence standards (Section 261). Thus, both figures share the underlying challenge—the reduction of the human factor in corporate life—but differ in their power structure and, above all, in the rigor of their safeguards. The truth is that the legislator has surrounded the more radical figure (the decentralized DAO) with greater precautions than the intermediate figure (the centralized automated company), when protective logic would advise at least similar treatment. III. Tension with the Foundations of the Corporate Legal Transaction III.1. The Cause (causa fin) of the Corporate Contract A particular mental habit permeates much of corporate legal analysis: the interpretation of its rules and principles as a closed, self-sufficient framework that excludes the rules of other areas of law. It is rarely remembered that companies, in their formation and operation, cannot infringe upon the constitutional guarantee of property, nor can they bypass the mandatory supplementary application of private law. This starting point, which we have sustained when studying the foundations of the corporate legal transaction (5), acquires unexpected relevance regarding the automated company, because it is precisely the logic of technological efficiency that threatens to set itself up as a new closed framework, indifferent to the general principles of contract law. One of the elements inherent in the foundational act of any company is the intention to enjoy the profits obtained by the enterprise, highlighted by the General Corporations Law (LGS) from its very first provision. Companies are formed and capital contributions are pledged to obtain something in return: participation in profits and, conversely, exposure to losses. That intentio iuris forms part of the underlying cause (causa fin) of the company contract, to the extent that the law renders null and void any provisions that exclude a partner from profits, as being incompatible with the essence of the transaction. However, the causa fin of the corporate contract is not exhausted by the expectation of profit. It also encompasses the willingness to combine efforts and resources for the achievement of a common goal, with the correlative expectation of participating in the business decisions that determine the fate of that investment. The affectio societatis, understood not as a mere psychological disposition but as a readiness to actively collaborate in the development of the enterprise, presumes the existence of management upon which the human will of the partners can, ultimately, exert an influence. What happens, then, when management no longer depends on human will, but on the execution of an autonomous algorithm? A partner who enters expecting to participate—even indirectly, through corporate governance—in the running of the business faces an entity governed by code, whose decision-making logic may prove as alien and impenetrable as that of a third party. This raises the first major question: Does the total automation of management distort the causa fin of the corporate contract, or merely modify it? The answer, we understand, allows for nuances. The automation of the execution of the corporate purpose—such as the algorithm entering into contracts, setting prices, or managing inventory—does not in itself alter the causa fin, just as the delegation of operational management to managers or hired third parties does not alter it. What can indeed frustrate the causa fin is the automation of structural decisions—establishing reserves, distributing dividends, strategic business orientation—without an effective channel remaining through which the will of the partners can be expressed and prevail. Here, the frustration of the contract's purpose acts as a legal limit to total automation: a limit that the Reform Bill, by failing to distinguish between the automation of execution and that of governance, completely overlooks. It should be emphasized, however, that automation, properly channeled, can strengthen the partner's position rather than weaken it. An algorithm whose decision criteria are predefined in the bylaws or in an auditable protocol drastically reduces the scope of opportunistic discretion by the director; the classic corporate conflict—the director managing for their own benefit and to the detriment of the partner—is mitigated when the rules of decision are transparent, verifiable, and automatically enforced.

III.2. Party Autonomy and its Constitutional Limits The Reform Bill is based, in its general conception, on strengthening party autonomy and contractual freedom: greater flexibility in statutory drafting, flexibilization of corporate types, and the expansion of the private dispositive sphere. The automated company is, in a way, an extreme expression of that philosophy: the freedom to organize the enterprise in the manner the founders deem most efficient, even dispensing with human intervention in operations. But party autonomy is not an absolute principle. As the best doctrine teaches, corporate rules do not constitute a closed system: above them stands the constitutional guarantee of property, and below them operates the supplementary application of private law, with its principles of good faith, abuse of rights, and protection of the weaker contracting party (6). The freedom to establish an automated company therefore meets insurmountable limits when its exercise compromises the property of minority shareholders or third parties. The specific risk is that the automated company may become a structure that systematically privileges technological efficiency over commutative justice among shareholders. Efficiency is a value, but neither the only one nor the supreme one. When algorithmic logic is oriented exclusively toward the optimization of a result (typically, the maximization of value for the ultimate beneficiary or the controlling shareholder), and there is no human counterweight to weigh conflicting interests, the property guarantee of the non-controlling shareholder is left exposed. The party autonomy of the founders cannot become an instrument for the silent confiscation of the rights of those who join later, or of those who, being in the minority, lack the power to configure the algorithm.

IV. Legal Personality and Veil Piercing. The Risk of Algorithmic Opacity

IV.1. Legal Personality as a Useful Fiction and its Limits The very formulation of the concept of legal personality implies a compromise between facts—the actions of human beings that produce legal effects attributable not to them directly, but to an ideal entity—and fiction. It is not that fictions are per se objectionable in an organizational technique such as law; what is ill-advised is elevating them to realities of the same category as facts in the physical world. In the line of Gervasio Colombres, faithful to Kelsen's thought, one should not equate "person" with "man": a legal entity is merely the personification of an order that regulates the conduct of several individuals, a common point of attribution for human acts determined by that order (7). This normativist understanding of legal personality—which we developed when studying the piercing of the corporate veil (8)—is key to understanding what makes the automated company disruptive. If a legal entity is a point of attribution for human acts, what remains of it when the acts attributed to it are not human, but algorithmic? The automated company pushes the abstraction of personality to its ultimate expression: an entity without permanent human organs, whose action is the output of a system. The fiction, which ultimately always rested on identifiable human conduct, runs the risk of becoming a second-degree fiction: a personification of acts that, in turn, are not directly attributable to any specific human being. This is the critical point. Legal personality was always a technique of attribution, not a mechanism for diluting liability. The law tolerated the fiction because, behind the entity, there were always human beings—partners, directors—who, in the event of abuse, could be reached by lifting the veil. The automated company threatens to break that chain: if the entity acts on its own (through the algorithm) and there are no permanent human directors, piercing the veil may reveal, behind the curtain, not a responsible human being, but another layer of code.

IV.2. The Automated Company as a Potential Shield of Impunity Based on the above, it is clear that the absence of robust mechanisms to identify those who exercise effective control can turn the automated company into a prime instrument for fraud (9). The scheme is familiar in its structure, though magnified in its capacity for harm because the profits of the operation flow to the ultimate beneficiaries—who remain hidden behind the technological facade—while the losses are attributed to a legal entity whose assets may be deliberately insufficient. The company is liable "with its assets," redundantly posits Section 14. But if those assets have been drained, or were never significant, asset liability is reduced to an empty formula. The Reform Bill reformulates, in Section 3, the doctrine of disregard of legal personality from Section 54, third paragraph, of the LGS. It allows disregarding the entity's personality when the company is used for extra-corporate purposes, to violate the law, public policy, or good faith, or to frustrate the rights of partners or third parties, "without requiring proof of fraudulent or harmful intent," with the effect of extending or shifting liability to the partners or controlling parties. Dispensing with such proof aligns with what we have maintained: piercing the veil does not require subjective blameworthiness, but rather the deviant use of legal personality or the need to give primacy to mandatory rules or the rights of third parties. From this perspective, the fact that an algorithm is the actor does not theoretically prevent the application of the doctrine, since the occurrence of an outcome objectively contrary to law or third-party rights is sufficient. The difficulty lies on another plane. Veil piercing requires identifying to whom the action should be imputed—the partner, the controlling party, or another company forming part of the same corporate enterprise—and it is this identification that algorithmic opacity and the absence of permanent human directors render difficult. Compounding this, Section 3 provides that the declaration of inapplicability "is of a restrictive nature," in line with the restrictive doctrines we have criticized. The doctrine is not inapplicable in theory—there will always be, at the origin, a human who configured the system for a purpose—but its practical operability is, in our view, compromised. The comparison with the DAO is once again revealing. For the DAO, the bill requires mandatory human representation (Section 260) and traceability with identification of owners in accordance with due diligence standards (Section 261). These requirements, precisely, are designed to prevent decentralization from turning into anonymity, and anonymity into impunity. The automated company, which can achieve analogous levels of depersonalization, lacks these safeguards. It is an omission that the legislator should correct prior to enactment: a corporate vehicle cannot be permitted without a legal mechanism for identifying the true beneficial owner and the party ultimately responsible for configuring the system.

V. The Management Body. A Conceptual Fracture

V.1. The Classical Paradigm The separation of ownership and management is the structural fact that explains the law of large enterprise. In the classical corporation, the owners of capital entrust management to a governing body—the board of directors—composed of human beings who act as a kind of agent of the company, subject to fiduciary duties of loyalty and diligence. Corporate management was always conceived as a function of human beings, insofar as it presumes attributes that only humans possess: judgment, discernment, ethical evaluation, prudent discretion. The standard of conduct imposed by law confirms this. Section 59 of the LGS requires directors to act "with loyalty and with the diligence of a prudent businessperson"; Section 91 of the Reform Bill, for its part, incorporates that standard—with nuances—by requiring them to act "with loyalty toward the company and with the individual diligence typical of a prudent businessperson." The "prudent businessperson" is not an automaton executing rules; the standard in question constitutes a pattern of conduct that presumes the ability to weigh circumstances, anticipate consequences, balance conflicting interests, and decide prudently in conditions of uncertainty. It is, strictly speaking, an anthropological standard: it measures the conduct of a reasonable and diligent human being placed in the position of the director.

V.2. Acts of AI Agents. The Problem of Legal Imputation At this point, the deepest conceptual fracture occurs. Artificial intelligence agents lack will in the technical-legal sense. They are not legal subjects: they cannot hold legal relations, incur obligations, or act as agents or proxies. Representation—whether voluntary or organic—presupposes a representative who is themselves a center of legal imputation: a person. An algorithm does not fit into any of the categories of the theory of representation, because it is not a person, nor does it have a will that can be imputed. Hence, the automated company poses a problem logically antecedent to corporate law. Before admitting that a company operates "through artificial intelligence agents," the legal order must define under what conditions the acts of an AI agent are legally valid and to whom they are imputed. Is the algorithm an instrument used by a human—such that its acts are imputed to the person who configured or operates it, as the act of an employee or tool is imputed to the principal? Or does the bill intend for its acts to be imputed directly to the company, without human intermediation, as if the system were a corporate organ? If the former, then there will always be a responsible human, and the concept of a company "without human resources" is actually a formulation that is as novel as it is devoid of practical utility. If the latter, a center of legal action is created without a responsible subject behind it, which is dogmatically impermissible (10). The truth is that the Reform Bill does not resolve this preliminary question. What is required, prior to the corporate figure, is a framework establishing the registration of algorithmic system operators, traceability standards that allow reconstructing which system made which decision and under what configuration, and the conditions of functional equivalence between the action of the AI agent and representative authority.

V.3. New Duties. Technological Supervision Fensore and Corbella have identified, in relation to the use of AI on boards of directors, a catalog of duties that enriches the standard of the prudent businessperson in the algorithmic era: the duty to properly select technological tools, to evaluate the risks associated with their use, to verify the quality and representativeness of the data used to train and feed the system, to implement human oversight over sensitive decisions, and to establish cybersecurity protocols (11). This catalog is entirely endorsable and points in the right direction: not replacing the director with a machine, but expanding their fiduciary duties to include technological supervision. Section 102 of the Reform Bill itself reflects this by referring to the "duty of configuration and supervision of the system and its results." However, all the aforementioned duties presuppose a human subject to exercise them: someone to select the tool, evaluate the risk, verify the data, and implement the controls. In the "pure" automated company of Section 14—the one operating "without human resources for its ordinary operation"—who exercises these duties? If the answer is "no one, because the company has no permanent directors," then the model is unfeasible, because there is no subject of the supervisory duties. If the answer is "the partners or a director who remains for configuration and supervision, though not for ordinary operation," then the figure is coherent, but its designation is misleading: it is not a company "without humans," but a company with humans relegated to the role of configuration and control, which is precisely the hybrid governance model. This second interpretation is, in our view, the only one compatible with the rest of the legal order (12).

VI. Liability of Directors and AI Agents

VI.1. The Current Liability Regime In few areas is the divorce between the good intentions of the law and the poor results of its application as marked as in actions for directors' liability. The intentio legis of the General Corporations Law was to ensure lawful and prudent management through an appropriate civil liability regime for directors. However, between restrictive legal doctrines and timid judicial practices, the scope of effective enforcement of the system has been reduced to almost nothing. Because directors are elected by the majority, and their appointment falls on the controlling parties themselves or people close to them, corporate liability actions are rarely brought, and when they do reach the courts, it is usually within the context of bankruptcy proceedings. The regime, in short, was already largely inoperative with human directors. Sections 59 and 274 of the LGS establish the joint, several, and unlimited liability of directors for the improper performance of their duties. Section 101 of the Reform Bill softens this: it configures it as subjective and individual, qualifies the director's obligation as one of best efforts (means), rules out their status as guarantor of corporate obligations, and limits joint and several liability strictly to liability toward the company. If the current regime was already difficult to enforce with respect to flesh-and-blood directors, automation exacerbates the problem to a critical point. When the harmful decision stems not from an identifiable director but from an algorithmic system, the evidentiary difficulties that already rendered liability illusory are multiplied.

VI.2. The Algorithm and the Proof of Negligence The law of tort and civil liability rests, in its classical structure, on the possibility of reconstructing a causal chain: identifying the conduct, the basis of liability (intent or negligence), the damage, and the causal link. Following these principles, directors' liability for breach of duty requires demonstrating that there was a negligent or disloyal action or omission. However, many contemporary artificial intelligence systems—particularly those based on deep learning—exhibit the phenomenon known as the "black box": their internal decision-making processes are opaque and non-interpretable, even to their own developers. The system produces an output, but the reasons why it arrived at it cannot be reconstructed in terms of an explainable logical sequence. This opacity conspires against the practical possibility of proving negligence. How can it be proved that a harmful algorithmic decision was the product of negligent configuration, and not an unpredictable and unavoidable outcome of the system? How do we distinguish culpable error from the margin of error inherent in any statistical tool? The burden of proof, which under the current regime already condemned liability actions to a lack of practical application, becomes an almost insurmountable barrier in the algorithmic context. Moreover, the Reform Bill incurs a contradiction worth emphasizing. On the one hand, it intends to subject the automated company to a regime of control and oversight. On the other, oversight cannot reasonably extend to the system's source code, which is protected by the developer's intellectual property and, absent a court order, shielded from the scrutiny of oversight bodies and the partners themselves. Yet that code is precisely what governs, decides, and executes. This creates an oversight system that cannot reach the very object it is supposed to supervise. Formal transparency coexists with substantive opacity.

VI.3. The Inadequacy of Corporate Asset Liability

As noted, Section 14 of the Reform Bill posits that the automated company "is liable with its own assets." We have already pointed out that this statement is, in itself, redundant because every legal entity is liable with its assets. The real question is what happens when those assets prove insufficient to redress the damages caused by the system, and there are no liable human directors whose personal assets supplement those of the company. In the traditional company, the entity's asset insufficiency is compensated for—at least partially, and where the liability regime actually works—by the personal liability of directors (unlimited, joint, and several in the current LGS; subjective, individual, and joint and several only toward the company under Section 101 of the Reform Bill). In the "pure" automated company, that second line of defense disappears (13). The result is an externalization of risk toward third parties and creditors. The benefits of algorithmic operation are captured by the ultimate beneficiaries, while the costs of its failures, to the extent they exceed the corporate assets, are borne by third parties. This asymmetry is unacceptable and calls for, as a condition for admitting the concept, complementary guarantee mechanisms: sectoral guarantee funds, mandatory liability insurance for algorithmic damages, minimum capital requirements proportional to the risk of the activity, or a strict liability regime for algorithmic products that reaches both the developer and the operator of the system (14). The simple declaration that the company answers with its assets, without the backing of real guarantees, is a promise of redress that in serious cases will go unfulfilled.

VI.4. The Business Judgment Rule in an Algorithmic Context The Reform Bill incorporates, in line with the evolution of comparative law, a version of the business judgment rule. A director who makes a business decision in good faith, based on sufficient information and through an appropriate procedure, is not liable merely because the decision turns out to be wrong. The rule protects entrepreneurial discretion and prevents judges from reviewing, with the benefit of hindsight, the risky decisions inherent in economic activity. Transposing this rule to an algorithmic context generates—at the very least—difficulties. Can an algorithm act "in good faith"? Good faith is a notion that presumes a subjective state—an honest belief in the correctness of one's actions—which a system entirely lacks. What constitutes "sufficient information" when data processing is opaque and one cannot verify what information actually influenced the decision? How is the adequacy of the "procedure" assessed when the procedure is a statistical model? The business judgment rule was designed to protect the prudent judgment of a human being; applied mechanically to an algorithmic decision, it risks becoming a safe harbor for unaccountability: it would suffice to invoke that "the system decided according to its parameters" to shield any harmful outcome. The business judgment rule can only make sense, in an automated company, when applied to the human decision to configure and oversee the system, not to the algorithmic decision executing it. This leads back, once again, to the hybrid governance model: the protection of business judgment covers the human who designs and controls, not the machine that executes (15).

VII. Shareholders' Rights and Digital Corporate Governance

VII.1. Economic Rights. Dividends, Reserves, and Capital Net equity accounts represent, in substance, shareholders' rights against the company: capital, reserves, and accumulated earnings are nothing other than the accounting expression of the shareholders' participation in the entity. As we have maintained when studying shareholders' rights over corporate assets (16), the policy on establishing reserves, capitalizing earnings, and distributing dividends directly affects the economic substance of those rights and constitutes one of the classic battlegrounds between majority and minority shareholders. The systematic retention of earnings—the creation of excessive or unjustified reserves—is one of the traditional avenues by which the controlling majority can strip the minority shareholder's right to dividends of its substance. In the automated company, this conflict takes on a special dimension. An algorithm could decide to establish reserves, capitalize earnings, or indefinitely defer dividend distributions without direct human intervention and without the deliberation that, at least formally, precedes these decisions in a traditional company. If the system's criteria are geared toward maximizing value for the controlling shareholder, the minority shareholder may find themselves deprived of their share in profits by a decision that emanates not from a deliberative assembly, but from a rule of code that they did not configure and cannot modify. The risk of dispossessing the minority shareholder—ever latent in dividend disputes—is aggravated when the decision is automated and removed from debate. Here, opportunity and danger coexist particularly closely. An algorithm whose distribution policy is predefined in the bylaws in a transparent and binding manner—"X% of the liquid and realized profits of each fiscal year will be distributed as dividends"—protects the minority better than an assembly deliberation dominated by the majority, because it removes the decision from the controller's discretion. Automating dividend policy can be a powerful tool for minority protection. Everything depends on who configures the rule and whether the minority has a voice in that configuration. If the rule is set unilaterally by the controlling party and the minority can only adhere to it, automation exacerbates the dispossession; if the rule is agreed upon, transparent, and enforceable, it prevents it. The Reform Bill, however, offers no guidelines in this regard.

VII.2. Governance Rights. The Shareholders' Meeting in a Company Without Humans A shareholder's governance rights—the rights to information, to speak, to vote, to challenge decisions—are exercised, in the classical framework, through the general meeting: the governing body that expresses the corporate will. The meeting presupposes a notice of call, deliberation, and decisions adopted by individuals. How does a shareholders' meeting operate in a company that lacks a human board of directors to convene it? Who renders accounts to the shareholders for management that was not human? Before whom is the right to information exercised, if there are no directors to provide information? The digitization of corporate bodies—remote meetings, electronic voting, digital corporate domiciles, which the bill otherwise admits and encourages—resolves the instrumental aspect of the problem, but not its core. The fact that a meeting is held by videoconference and voted on using digital signatures does not change the reality that there must be someone who calls the meeting, provides information, submits management to shareholder approval, and answers for it. Digital corporate governance is a valuable modernization of procedural forms. However, the automated company poses a problem of substance: the potential disappearance of the accountable subject. Without a director—even one reduced to the role of configuration and oversight—to call the meeting, inform the shareholders, and answer to them, governance rights become rights without an addressee, prerogatives that cannot be enforced against anyone. This also confirms that the automated company cannot entirely dispense with a human point of attribution.

VIII. Comparative Law and International Experiences Comparative experience, while still nascent, offers valuable lessons and warnings that the Argentine legislator would do well not to ignore. The State of Wyoming, in the United States, was a pioneer in legally recognizing DAOs in 2021, providing them with a specific legal form within its limited liability company regime (17). The experience, however, soon revealed its limits: the obligation to identify beneficial owners, imposed on limited liability companies by federal legislation (Corporate Transparency Act), proved impracticable for organizations with thousands of anonymous participants scattered around the world. The tension between the decentralized and anonymous logic of the technology and the identification requirements typical of corporate and tax law forced the State to review its approach and introduce, in 2024, additional legal structures and adjustments to alleviate these shortcomings (18). Malta, which branded itself as the "Blockchain Island" with early and ambitious legislation to attract the crypto-asset industry, ended up on the Financial Action Task Force (FATF) "grey list" due to deficiencies in anti-money laundering frameworks, including inaccuracies in beneficial ownership information. While the inclusion was not exclusively due to its crypto regime, the Maltese case illustrates the reputational and economic cost of a regulatory framework that prioritizes attracting investment and innovation without ensuring identity and integrity controls (19). The European Union, for its part, has taken a different and, in our view, more prudent path by addressing the regulation of artificial intelligence as such—with its Artificial Intelligence Act (Regulation (EU) 2024/1689, known as the AI Act)—prior to and independently of its incorporation into corporate law (20). The European approach, based on classifying systems according to their level of risk and imposing obligations proportional to that risk, provides a general framework upon which sectoral regulations—including corporate law—can subsequently rely. It is plain to see that the order of the steps affects the outcome: first, the legal status and obligations of the AI system as an agent are defined, and only then is it incorporated into the various fields of law. What emerges from these experiences is that the corporate regulation of automation, without a prior framework regulating artificial intelligence as a legal agent and imposing the identification of responsible parties, creates insurmountable gaps. Wyoming and Malta show the costs of haste; the European Union shows the virtue of the correct sequence. The Reform Bill, by creating the automated company without that prior framework, comes dangerously close to the former model.

IX. By Way of Corollary The automated company constitutes an understandable—and in its intention, valuable—response to the need to adapt corporate law to the digital economy. The automation of business activity is a real, irreversible phenomenon that carries genuine opportunities: cost reduction and a decrease in opportunistic managerial discretion, greater speed and traceability of transactions, real-time accounting, tokenization and democratization of access to investment, and new business models. From a decidedly favorable perspective, it has even been argued that this vehicle represents an opportunity for Argentina to lead the region in the legal recognition of the algorithmic economy and to attract investment and innovation that currently lack an enabling regulatory framework (21). Nothing said in this work denies these opportunities, but their realization depends on building the concept upon solid legal foundations, and it is precisely there that the Reform Bill, in its current wording, proves premature and insufficient. Indeed, a company without an identifiable, real, and accountable human point of legal attribution is inconceivable. Legal personality was always a technique for attributing human acts, not a mechanism for dissolving responsibility. A company that acts on its own, through an algorithm, without any human being answering in the last instance, contradicts the very nature of the institution. And the mere declaration that the company "is liable with its assets" is insufficient without complementary guarantee mechanisms—guarantee funds, compulsory insurance, minimum capital proportional to risk, strict liability for developers and operators. From this perspective, it is clear that the corporate concept requires prior legal definitions that the Reform Bill omits and that exceed the scope of the General Corporations Law: the legal status of the acts of an artificial intelligence agent (to whom they are imputed and under what conditions they are valid), a registry of algorithmic system operators, mandatory traceability standards, and a liability regime for damages arising from algorithmic products and decisions. Without resolving these issues, the automated company is an edifice without solid legal foundations. The international experiences reviewed confirm the risks of regulatory haste. Wyoming and Malta show that recognizing a legal form before resolving the identification of responsible parties exposes the system to opacity, deficiencies in anti-money laundering compliance, and supervisory difficulties. The European Union, by contrast, illustrates the virtue of the inverse sequence: regulating artificial intelligence as an agent first, and subsequently integrating it into different fields of law. The Reform Bill would do well to take inspiration from this second model. In our view, the preferable model is not the replacement of the human director with an algorithm, but rather hybrid governance: artificial intelligence as a powerful tool for strategic and operational support, but with ultimate responsibility resting on human beings who configure, supervise, and answer for it. This model is the only one compatible with the structural principles of corporate law. The tension between Section 14—which suggests depersonalization—and Section 102—which assumes responsible human directors—should be resolved in favor of the latter, making explicit that even an "automated" company retains a human core of configuration, supervision, and liability. It is telling that the bill has been more demanding with the DAO—imposing mandatory human representation (Section 260) and traceability with identification of owners (Section 261)—than with the automated company, on which it imposes none of these precautions (22). This asymmetry should be corrected by extending to the automated company, at the very least, the requirement of a responsible human representative, as well as standards of traceability and beneficial owner identification. Ultimately, it is understood that true innovation does not consist in replacing the director with an algorithm, but in building a regime that combines technological efficiency with legal responsibility, transparency, and the protection of the rights of shareholders, creditors, and third parties. Corporate law is a system for protecting interests and ensuring commutative justice among those who participate in the enterprise and those who interact with it. Automation will be progress if it serves these ends, and a regression if it sacrifices them on the altar of efficiency. It is up to the legislator—and the legal doctrine that must guide it—to ensure the former and prevent the latter.