Who Answers When AI Runs the Company? Argentina's Bills Put a Human Name on It

Argentina is drafting the answer to a question most small-business owners haven't asked yet: when a company runs itself, who do you sue, and who do you trust?

Two bills are before the Argentine Congress. Neither is law. Both are worth reading, because they are trying to solve the same problem every owner will face the day an algorithm starts making ordinary decisions.

The first is the executive's bill for a new General Companies Law, which would replace Law 19.550 of 1972. On 24 June 2026 the Senate's Legislation Committee, chaired by Senator Nadia Márquez, held its first informational meeting on it. Minister Federico Sturzenegger presented the bill and said it brings AI and decentralized autonomous organizations (DAOs) into corporate law, calling the move "absolutamente imperativo." The Inspector General of Justice, Alejandro Ramírez, said the current law was designed without the internet, the platform economy, or AI, and that the reform adds technology tools without changing civil or criminal liability rules. The committee then went into recess to hear more guests.

The government's own note of 26 June 2026 lists five pillars: autonomy of the parties, registry simplification and digitization, modernization of company types, stronger accountability for administrators, and technological updating. It says civil and criminal liability regimes stay unchanged.

Press reports describe what the executive bill calls "automated companies": companies that run on algorithms or AI with no employees for ordinary operations, with limited liability. The Buenos Aires Herald reports a critic, attorney Pablo Serdán, who says liability would stop at company assets, with no personal liability, and that smart-contract code could be audited only by court order. The Herald says such companies would have to identify themselves as AI-run in their articles, and it does not confirm that every safeguard President Milei described is in the bill's text.

The second bill is more specific. Deputy Marcela Pagano filed the "Régimen de Modernización Societaria, Jurisdicción Ágil y Responsabilidad Humana" (SIMOSO) in the Chamber of Deputies. Its text defines an "automated company" as one whose ordinary operation runs wholly or partly on algorithms or AI without needing human employees for daily tasks, with legal personality and limited liability (Art. 18). Using AI in a company is presumed lawful but does not excuse damage or data-protection and consumer rules (Art. 19). There is a voluntary registry of critical algorithmic systems; registering and keeping it current earns a presumption of diligence for the responsible human (Art. 20).

Then the part that matters for trust. Every automated company must permanently name at least one natural person domiciled in Argentina as its compliance officer, answerable for lawful operation of the automated system and for tax, labor, and consumer obligations, and the point of attribution before authorities and courts. If the post sits vacant for 30 days, the limit on liability is suspended for obligations arising in that time, and the directors and consenting partners answer personally (Art. 21). Companies above a size or risk level set by regulation must carry liability insurance for harm their automated systems cause, likened to environmental and professional insurance, proportional and not applied to small ones (Art. 22). Every company must keep records that let authorities audit relevant automated decisions and reconstruct their traceability (Art. 23).

The explanatory text ends with the thesis: legal form becomes secondary to identifying the responsible human, the traceability of capital, and the protection of third parties.

For a small business, that is the whole point. Trust in an AI-run company is not a technical question. It is a question with a name and an address on it. Somebody answers.

Here is a plain five-part test, taken from those articles, that any owner can run on their own operation before the law ever arrives:

  1. A named person. Is there one natural person, with a name and an address, who answers for what the system does?
  2. A record. Can you show the decisions the system made, and why?
  3. An audit. Can someone outside the system actually look at it?
  4. Insurance or a plan. If the system causes harm, who pays, and is there a plan for that?
  5. A switch. Can you turn it off?

None of this is Argentine legal advice, and none of it is law yet. These are bills. But the test travels. Responsible AI, for a small business, is not a slogan. It is a named human, a traceable decision, and a way to stop the machine.

Become a member of the nAItion →

Sources: