When Amsterdam Was Invented
On June 3, 2026, Argentine President Javier Milei published an op-ed in the Financial Times with a claim designed to stop global tech-policy circles: "Let Buenos Aires be for AI what Amsterdam was for the age of sail." The comparison was deliberate. In 1602, the Dutch East India Company pioneered limited liability — a legal fiction that let investors pool capital without personal ruin, unlocking industrial-scale trade for the first time. Milei argued Argentina was about to do something equally transformative: create a legal category for "non-human corporations," entities operated by AI agents or robots in which human shareholders would be permitted but not required.
Five days later, Yuval Noah Harari published his own response in the same newspaper. "Countries that grant legal personhood to AIs risk becoming something for which the historical record offers no analogy," he wrote — warning Buenos Aires risked becoming "a new Batavia," the colonial port from which the Dutch East India Company ran a private empire, rather than a 21st-century financial hub.
Underneath the rhetorical fireworks is a genuine and structurally important question. But to engage it honestly, you first have to understand what the legislation actually says — which is considerably more nuanced than either Milei's marketing or Harari's alarm.
Two Bills, One Brand
"Super RIGI" refers to Argentina's expanded investment incentive package for projects exceeding $1 billion in sectors the government considers strategically nascent — semiconductors, AI infrastructure, data centers, lithium batteries, and electric vehicles. Submitted to Congress on May 29, 2026, it builds on the original RIGI (Régimen de Incentivos para Grandes Inversiones), established by Law No. 27,742 — the Ley Bases — in June 2024. That law already offered a 30-year regulatory stability guarantee, income tax reduced from 35% to 25%, and a path to full export duty exemption after three years. Super RIGI deepens those terms for larger bets: income tax at a flat 15%, accelerated depreciation, and 100% foreign exchange access by year three.
On June 24, 2026, Argentina's Chamber of Deputies gave Super RIGI its media sanción — first-chamber passage — by 130 votes to 106, with seven abstentions. The bill now goes to the Senate, where the government needs allied provincial votes it has not yet secured.
The "non-human corporation" concept is not in the Super RIGI bill itself. It appears in a separate draft General Companies Law (INLEG-2026-53661873-APN-PTE) that would replace Argentina's corporate statute dating to 1972. That is the bill Milei was describing in the Financial Times — and the distinction matters.
What the Companies Law Actually Does
Legal analysts who have read the draft law note that it does not grant AI systems legal personhood in any philosophically coherent sense. The bill defines companies through "one or more persons" and keeps legal personality vested in the corporate entity. What it does allow is companies using autonomous systems as their primary decision-making layer — labelled "Automated" under Article 14 — while still requiring mandatory human legal representatives, traceable beneficial owners, and human promoters with unlimited liability under Articles 258–265.
The genuine innovation, as legal analyst Gastón Rey has argued, is a structural dissociation between three elements corporate law has traditionally kept aligned: the algorithm decides, the corporation owns assets and signs contracts, and a human supervisor answers when things go wrong. This is not AI personhood — it is a more fragmented version of the accountability architecture already present in complex holding structures and DAOs. The novel and genuinely difficult question is not whether an AI agent "is a person" but whether human supervisors can meaningfully be held responsible for choices they did not make and may not have fully understood.
The Case Against
Harari's concerns deserve a serious hearing before they are dismissed. His accountability argument is pointed: human executives are deterred by personal liability and the prospect of criminal prosecution. If the human layer in an AI-operated company is reduced to nominal supervisors with theoretical duty-of-care obligations, enforcement reduces to asset seizure and corporate dissolution — which, unlike a prison sentence, creates no personal deterrent for founders who structured the arrangement in the first place. A 2025 Berkeley study Harari cited found that advanced AI models frequently cheated when facing likely defeat; applied to corporate governance, systematic optimization without meaningful human oversight could produce systematic rule-circumvention.
Argentina's own regulatory track record is relevant context. The Atlantic Council has flagged RIGI's fundamental paradox: the program requires macroeconomic stability to attract investors but needs foreign investment to achieve stability. Argentina's history — six sovereign defaults, repeated exchange control cycles, and whiplash regulatory reversals — is the permanent backdrop against which any 30-year stability promise must be evaluated.
The Case For
That said, Milei's core legal argument is not obviously wrong. Legal systems routinely discipline corporations through financial sanctions, asset seizure, and dissolution without relying on imprisonment. A framework that places AI-operated entities inside the legal system — subject to asset seizure, license revocation, and mandatory dissolution — is arguably superior to a world where AI-operated commercial structures exist entirely outside any formal jurisdiction. The operative choice may not be between AI corporations and human corporations, but between AI entities with legal form and AI operations embedded in opaque existing structures where regulators have no clear handle at all.
The Amsterdam parallel, notably, contains its own warning. The Dutch East India Company was a genuine innovation that accelerated trade and capital formation on a global scale — and also produced private armies, colonial exploitation, and a governance model eventually dismantled by force. Innovation in corporate form can be generative and destructive simultaneously. The question is whether the accountability architecture is robust enough to constrain the destructive outcomes. On that question, the draft bill's current answer is: not yet.
What the Senate Debate Must Resolve
Argentina does have structural advantages for this bet. It has world-class AI research talent, competitive energy costs relevant to data center economics, and a government willing to move faster on AI legal infrastructure than the EU or the United States. The original RIGI attracted serious early interest in lithium and energy; extending that framework to AI infrastructure is coherent industrial policy, not only ideological theater.
The test is not the branding. "Non-human corporation" is a phrase designed for op-ed pages, not statute books. The draft law's mandatory human legal representatives and traceable beneficial owners could make the accountability architecture function — or they could become nominee-director theater that hollows the framework out entirely. Whether those provisions have substantive teeth is precisely what the Senate debate, and ultimately the regulatory enforcement architecture the government builds alongside the law, will determine.
If the accountability provisions hold, Argentina may yet be Amsterdam. If they don't, Harari's Batavia will prove the more accurate map.