Can A.I. competently run a company without human oversight?
On 29th May, Argentina’s government sent a bill to Congress that would amend a long-standing corporations law (since 1972) to recognise a new kind of legal entity: the “non-human corporation” – a company that can be operated entirely by AI, where a human shareholder is permitted but no longer required. President Milei and his deregulation minister set out the logic in a Financial Times op-ed, reaching back to the Dutch East India Company of 1602 to argue that limited liability is the legal innovation each technological era demands, and that the AI era now needs its own.
Most people who saw the headline last week filed it under science fiction and scrolled on. They shouldn’t have; it is a real bill, in a real legislature, and other governments are reading it closely. The debate has triggered conversations. The problem is that almost everyone is having the wrong one.
So far, the commentary has fixated on the philosophical and existential questions like “how human can A.I. be?” and “are we entering an A.I. state?”. Yuval Noah Harari warned in his own FT response that Argentina risks becoming a haven for regulatory arbitrage. These are serious questions, but they’re mostly situated in the abstract.
There’s a more pertinent, practical question we need to ask to determine whether an A.I. company could actually function in practice.
A corporation is not a philosophical object. It is a thing that employs people, holds assets, signs contracts and – more often than not – operates across borders. Remove humans from the org chart and you leave certain areas of your business vulnerable. In particular, global mobility and cross-border operations – areas which require meticulous attention to detail and careful relationship management. Two things A.I. is not (yet) able to deliver on as well as a human being can.
Consider these three questions and whether a non-human employee would be equipped to deal with them to a sufficient standard:
Where is a tax-resident? Residence, in most jurisdictions, hangs on where a company is managed and controlled, by the people who call the shots. Remove those people and the test has nothing to go on. Is the entity resident where its servers run, where its agent happens to act, where it was incorporated, or nowhere at all? “Nowhere” is not a neutral answer. It is the foundation of the most aggressive tax structure ever conceived.
Who carries the duty of care? Picture a non-human corporation, incorporated in Buenos Aires, that deploys a workforce across fifteen countries; engineers on site visits; staff working remotely; contractors crossing borders. Every one of those movements triggers obligations: work authorisation, social security coordination, posted-worker notifications, a duty of care if something goes wrong. Those obligations do not evaporate because the employer is software.
And what happens when things go wrong? Because it will, inevitably. Maybe an employee is detained at a border on the wrong status, or a tax authority opens an enquiry into a permanent establishment nobody declared.
With every problem that arises, the same thing is required: a person who picks up the phone, or answers the email. Limited liability never meant no liability – it means a human principal should always stand behind the entity with defined exposure. Milei’s own Dutch East India Company understood that. Removing the principal entirely does not admonish accountability.
No border officer, tax inspector, or court has ever accepted “the software decided” as a defence.
This is the real risk in the Argentine proposal, and it is far more concrete than the personhood debate. It is the prospect of an entity that can operate a workforce everywhere while being accountable nowhere.
This takes us to the thing worth taking from all of this, whether or not Argentina’s bill ever becomes law. The more of corporate life we automate, the more valuable the one thing automation cannot supply becomes: judgement and accountability that holds up across borders.
We know that software is becoming increasingly adept at assessing risk. What we risk losing is the willingness and the standing to answer for the consequence, in the right jurisdiction, to the right authority, when it matters. In a world of non-human corporations, this becomes a scarce asset.
Argentina will not be the last country to try this; the incentives facing every government with a shrinking tax base are too strong. More of these entities are coming, in more places. The organisations that thrive alongside them will not be the ones with the cleverest automation. They will be the ones who never lost sight of where the human has to stand – between the automation and the consequence – because that is the one position a border, a tax authority or a court will always insist someone occupies.
The robots might be able to run the company, but they cannot cross the border for you.
Discover more global trends, mobility and travel regulation updates and company news.
The European Union’s new digital border control system, the Entry/Exit System (EES), is now fully operational across all Schengen Area borders as of 10 April 2026. This is a significant change in how British citizens and other non-EU nationals will enter and exit the Schengen Area, with major implications for business travel, compliance and border processes.
The European Union’s new digital border control system, the Entry/Exit System (EES), is now fully operational across all Schengen Area borders as of 10 April 2026. This is a significant change in how British citizens and other non-EU nationals will enter and exit the Schengen Area, with major implications for business travel, compliance and border processes.
Centuro Global is proud to announce that we have won the ‘Best Business Travel Management’ award at the Forum for Expatriate Management’s (FEM) Americas Expatriate Management & Mobility Awards 2026 (EMMAs).
Discover how we can help your team with immigration, business travel and global expansion.
Book a call
