No One to Call: What Milei's AI Corporation Proposal Means for Australian Procurement Risk

Argentina wants to rewrite their company law to let artificial intelligence own and operate businesses with no human principal required. It might sound like the background setting of a facilitating sci-fi movie; for procurement professionals, it is a third-party risk scenario worth taking seriously right now.

‍

The proposal

On 29 May 2026, Argentina's President Javier Milei submitted legislation to Congress to create a new category of corporate entity: the ‘sociedad automatizada’, or ‘non-human corporation.’ The draft bill amends Argentina's general corporations law, Law 19.550, which has been in force since 1972.[1]

In a Financial Times opinion piece co-authored with Deregulation Minister Federico Sturzenegger and published on 3 June 2026, Milei outlined three pillars underpinning the reform:[2]

1. Wholly unregulated AI. Milei committed to keeping artificial intelligence "free to be developed without the deadly hand of premature and poorly understood regulation."

2. A new corporate category. "These are entities operated by AI agents or robots," Milei stated. Human shareholders may participate, but are not required.

3. A competitive fiscal environment. A low corporate tax rate, with mandatory beneficial-owner disclosure for any human participants. Blockchain-based DAO structures are explicitly accommodated.[3]

"Where these systems exercise independent judgment in unpredictable environments, their actions entail real risks. Limited liability is not a luxury for such entities; it is a precondition for their existence." — President Javier Milei, Financial Times, 3 June 2026

Milei frames Buenos Aires as potentially becoming what Amsterdam was for the age of sail, invoking the 1602 founding of the Dutch East India Company as a historical parallel.[2] The response from outside Argentina was swift. Historian Yuval Noah Harari published a rebuttal in the Financial Times on 8 June 2026, warning that granting legal status to AI-operated entities would hand them access to financial, economic, and political systems, with no historical analogy for what a country might become under such arrangements.[4]

‍

The accountability gap at the centre of it all

The most commercially significant aspect of the proposal is not the novelty of the concept. It is the deliberate removal of a human principal from the legal chain. Under the proposed structure, an AI-operated entity can sign contracts, hold assets, and transact commercially, with no natural person required to be legally accountable for its decisions.[5]

AI Weekly, which tracks the legislative proposal in depth, noted that "the deregulation-first design removes the liability anchor: if an AI-operated entity causes harm, there is no human principal legally required to answer for it." The same analysis flagged a specific arbitrage risk: shell entities using minimal AI wrappers could incorporate in Argentina as non-human corporations to insulate human controllers from legal accountability.[6]

This is not a distant hypothetical. Existing contract law in most jurisdictions, including Australia, generally requires parties to hold legal personhood. When an AI agent acts autonomously across multiple systems, establishing who granted authority, whether that authority was binding, and who bears liability can become genuinely contested.[4]

‍

Where Australia sits right now

Australia does not have a comprehensive AI Act. The National AI Plan, released by the Department of Industry, Science and Resources in December 2025, explicitly rejected standalone AI legislation. The Government's stated position is to "continue to build on Australia's robust existing legal and regulatory frameworks," relying on existing technology-neutral laws and voluntary standards.[7]

The Australian AI Safety Institute was allocated AUD 29.9 million to launch in early 2026. Regulatory accountability remains fragmented across multiple bodies: ASIC, the OAIC, the ACCC, APRA, and the TGA, each applying their existing mandates to AI use cases within their sectors.[8]

In April 2026, APRA issued a direct letter to the banking, insurance, and superannuation sectors, calling for a step-change in how AI-related risks are managed. The letter specifically called out procurement as a risk domain, stating that AI risks "cut across multiple domains at regulated entities. This includes operational risk, cyber and information security, data governance, model risk, change control and release management, legal and regulatory compliance, privacy and conduct risk, procurement and third party dependency."[9]

APRA also observed that boards were showing an over-reliance on vendor presentations and summaries, without sufficient examination of key AI risks such as unpredictable model behaviour and the impact on critical operations.[9] That observation applies well beyond regulated financial entities.

Australia's Privacy and Other Legislation Amendment Act 2024 introduced reforms requiring entities to notify individuals when substantially automated decisions significantly affect their rights or interests. Those reforms come into effect in December 2026.[10]

‍

What this means for third-party supplier risk

For procurement professionals, the Argentina proposal surfaces a category of supplier risk that most third-party risk frameworks are not built to handle: a supplier with no identifiable human accountable party.

Australian procurement operates under an assumption, mostly implicit, that behind every supplier there is a legal person who can be held accountable. That person may be a director, a sole trader, a trustee, or a shareholder. The chain may be complex, but it exists. The non-human corporation breaks that chain by design.

Even if Australian law does not adopt equivalent structures, the risk is already forming at the edges. AI-native vendors are proliferating rapidly. The lines between a company that uses AI and a company operated by AI are already blurring. MinterEllison, writing in March 2026 on the legal complexities of acquiring AI companies, identified third-party AI dependencies and liability exposure as creating "concentration risks and novel accountability questions that acquirers must evaluate through tailored due diligence."[10]

The due diligence questions that matter now

The following questions are not speculative. They are relevant to any organisation procuring AI-enabled goods or services from a supplier that is either AI-native or heavily AI-dependent:

• Who is accountable? Can you identify a named natural person or recognised legal entity that carries liability for failures, errors, or harms arising from the AI component of the goods or services supplied?

• Where is the entity incorporated? If a supplier is incorporated in a jurisdiction with no mandatory human oversight requirements, does your contract still bind a party who can be compelled to remedy a breach?

• What does the contract actually cover? Hall and Wilcox, writing in 2026 on AI procurement in Australia, noted that poorly defined functionality or deliverables in AI contracts can lead to misaligned expectations and disputes about performance.[11] If the AI component of a service is undefined, so is the liability trigger.

• Where does the data go and who controls it? Australia's OAIC has issued guidance on AI and privacy. Accountability does not shift to the vendor under Australian law. If a third-party AI tool produces a discriminatory or unlawful outcome, the procuring organisation bears the regulatory exposure.[12]

• What are the operational dependencies? If a supplier's AI system fails, halts, or behaves unpredictably, what is the contractual mechanism for remedy? Many standard service agreements were not written with autonomous AI decision-making in mind.

‍

The case for and the case against

Argentina's proposal is not without a coherent commercial logic. Limited liability was itself a radical legal innovation when it emerged in the 19th century. It unlocked capital formation at scale by separating investor risk from enterprise risk. Milei is making a structural argument: that AI-operated entities, acting in unpredictable environments, require an equivalent legal scaffold before meaningful deployment is commercially viable.[2]

The argument has surface merit. An AI agent that signs contracts, manages logistics, and processes invoices does face a genuine legal identity problem under current frameworks. Forcing that activity into a human-controlled corporate shell may simply obscure the actual decision-maker rather than create meaningful accountability.

The argument against is more compelling. Harari's rebuttal in the Financial Times identified the core danger: granting legal personhood to AI corporations gives them an all-purpose key to access financial, economic, and political systems, with no corresponding accountability to citizens, regulators, or courts in the way that human-controlled entities are ultimately accountable.[4] The Heritage Foundation's Index of Economic Freedom shows Argentina moving up 20 positions in both 2024 and 2025, reflecting Milei's deregulatory agenda.[13] That momentum is real. The risk is that deregulation in this area is not the same as deregulation in others: the absence of a human in the accountability chain is a structural feature, not an administrative simplification.

For Australia, the pros and cons of adopting an equivalent structure break down as follows:

Potential benefits

• A clear legal identity for AI-native businesses operating autonomously, reducing the ambiguity that currently forces those entities into ill-fitting corporate structures.

• A framework that could attract AI-driven investment and incorporation, provided Australia pairs entity recognition with proportionate governance requirements.

• Regulatory clarity for procuring organisations, if entity recognition came with mandatory disclosure obligations that answered the due diligence questions above.

Significant risks

• Accountability erosion in supply chains, where the removal of a mandatory human principal creates gaps that existing contract, consumer, and privacy law cannot fill without substantial reform.

• Incorporation arbitrage, where entities use the structure to insulate human controllers from liability while retaining the economic benefits of AI-operated commerce. Australia has no appetite to become a haven for liability laundering.

• Regulatory fragmentation risk, compounded by Australia's existing multi-regulator approach to AI. Adding a new entity class without a single supervising authority would widen the gaps.

• Procurement exposure at scale, particularly for government and enterprise buyers who are procuring from AI-native suppliers with limited visibility into beneficial ownership, decision-making authority, and liability chains.

‍

What procurement professionals should do

Australia will likely not legislate a non-human corporation category in the near term. The National AI Plan reflects a deliberate preference for principles-based governance over structural reform. That is a defensible position, but it places the burden of managing supplier risk squarely on a procuring organisation.

The practical steps are not complex, but they require deliberate action:

1. Update your supplier onboarding to ask for AI disclosure. Any supplier where AI is material to service delivery should be required to identify the human or entity accountable for AI-related decisions and failures.

2. Check your contract templates. Standard service agreements rarely address autonomous AI decision-making, liability allocation for AI errors, or remediation obligations when AI systems behave unpredictably. This is a gap worth closing before a dispute opens it for you.

3. Understand your regulatory exposure. Under Australian law, accountability does not shift to the AI vendor when a third-party tool produces an unlawful or harmful outcome. Your governance obligations follow the decision, not the system that made it.[12]

4. Watch the Argentina experiment. Whether Milei's legislation passes and how it operates in practice will provide early evidence on whether AI-operated entities behave as coherent commercial counterparties. That evidence will matter for every jurisdiction that follows.

5. Align with APRA's direction even if you are not a regulated entity. APRA's April 2026 letter sets out the risk domains where AI governance must improve. Its framing of procurement and third-party dependency as a named risk category is a signal worth taking seriously across all sectors.[9]

‍

The future state: flags of AI convenience

There is a well-documented precedent for what happens when a jurisdiction offers minimal regulation, low costs, and concealed ownership as a package deal to attract commercial activity. The maritime sector has lived with it for decades. It is called the flag of convenience system, and its consequences for supply chain integrity are instructive.

A flag of convenience vessel is one that flies the flag of a country other than the country of its true ownership. The International Transport Workers' Federation (ITF), which has tracked and campaigned against the system for over half a century, describes it as enabling shipowners to choose the legal and regulatory regime that applies to their operations, selecting jurisdictions with lower costs and standards and limited enforcement capacity.[14] FOC ships now account for approximately 30 per cent of the entire global merchant fleet.[15]

The consequences are not abstract. The ITF documents four distinct categories of harm that the FOC system enables systematically:[14][16]

• Labour and welfare exploitation. FOC ships routinely pay below-standard wages, impose excessive working hours, and provide inadequate rest, food, and sanitary conditions. Crew abandonment, where seafarers are stranded without pay in foreign ports, is disproportionately concentrated in FOC-flagged vessels.

• Tax avoidance. Cheap registration fees, low or zero corporate tax, and offshore ownership structures allow beneficial owners to extract commercial profits while avoiding the tax obligations of their home jurisdictions. The ITF has explicitly described FOC registries as a systemic vehicle for tax avoidance.[17]

• Safety and environmental evasion. Many open registry states lack the enforcement capacity or political will to apply international safety and environmental standards consistently. The Seafarers International Union has identified FOC registries as enabling a growing network of vessels, the so-called dark fleet, that routinely evades sanctions, safety protocols, and environmental regulations through permissive flag states and obscured ownership structures.[18]

• Concealed ownership. Shell companies layered through multiple jurisdictions routinely obscure the beneficial owner of FOC-flagged vessels, fragmenting accountability and making enforcement practically impossible. Ownership anonymity, as Context News reported in August 2025, conspires with weak flag state enforcement to fragment responsibility across the entire chain.[19]

The parallel to Argentina's non-human corporation proposal is not superficial. It is structural. Milei's three pillars, no regulation, a novel corporate category, and a low tax rate, are the maritime flag of convenience model, rewritten for the AI era. Call it what it is: flags of AI convenience. The mechanism is different. The outcome is the same.

Call it what it is: flags of AI convenience. The mechanism is different. The outcome is the same.

‍

In the maritime sector, a shipowner registered in Panama or Liberia with a crew sourced from the Philippines and cargo destined for Europe could evade the labour laws of every jurisdiction involved, because no single national regulator held clear authority. In the AI sector, a non-human corporation incorporated in Argentina, operating AI agents that source goods, manage logistics, and process payments across Australian supply chains, could sit beyond the reach of the Australian Consumer Law, the Privacy Act, the Fair Work Act, and the Corporations Act simultaneously. No single regulator has jurisdiction over an entity with no mandatory human principal and no requirement for domestic incorporation.

The flags of AI convenience risk maps directly across each of the four FOC harm categories:

• Labour and welfare exploitation translates, in the AI context, to supply chains operated by autonomous systems that make sourcing decisions with no human oversight of the labour conditions embedded in those decisions. An AI-operated procurement entity has no ethical obligations and faces no reputational consequences. There is no person to hold accountable when a supplier of convenience sources from exploitative labour markets.

• Tax avoidance translates directly. A non-human corporation incorporated in a low-tax jurisdiction, transacting commercially across borders, generates revenue that flows to beneficial owners who need not be disclosed in the contracting jurisdiction. Australia's existing transfer pricing and thin capitalisation rules were written for human-controlled entities. A flags of AI convenience structure tests every assumption those rules make.

• Safety and regulatory evasion translates to AI systems that make consequential decisions in procurement, in financial services, in infrastructure, without being subject to the conduct obligations that Australian law imposes on human decision-makers. The OAIC's guidance is clear that accountability does not transfer to the vendor.[12] It does not, however, extend that accountability to an entity with no human principal at all.

• Concealed ownership translates with the addition of a new and more potent mechanism. In the maritime sector, shell companies obscure human owners. Under a flags of AI convenience structure, the absence of a mandatory human owner is a feature of the legal architecture, not a concealment technique layered on top of it. The opacity is baked in by design. It is also entirely lawful.

‍

What the maritime experience tells us about the likely trajectory

The FOC system was not designed to cause harm. It emerged from legitimate commercial pressures: competitive shipping markets, uneven international regulation, and the genuine need for capital formation in the post-war era. The first modern open registry, Liberia's, was created in 1949 as a result of a US-based commercial and political initiative.[20] The harms accumulated over decades as the system was exploited at scale, governance did not keep pace, and enforcement remained fragmented across dozens of jurisdictions.

The ITF's response took half a century to develop meaningful traction. Its collective bargaining agreements, which now cover approximately 47 per cent of all FOC vessels, represent the only internationally recognised enforcement mechanism for minimum standards on FOC ships.[15] They exist because no single national regulator could act alone, and because the commercial incentive to exploit the system consistently outpaced the political will to reform it.

The flags of AI convenience equivalent will move faster. The maritime sector operates physical assets that can be inspected, detained, and tracked. AI-operated entities operate in digital environments where the equivalent of a port state inspection does not yet exist. If Argentina's legislation passes and is replicated by other jurisdictions competing for AI investment, and the competitive pressure to do so will be real, the window for establishing governance frameworks that prevent the FOC pattern from repeating narrows quickly.

For Australian procurement professionals, the practical conclusion is the same one that responsible maritime buyers reached eventually: you cannot rely on the flag to tell you what the ship actually is. You have to look behind it. Supplier due diligence on AI-enabled and AI-operated vendors needs to ask who ultimately controls the system, where liability actually sits, what labour and ethical standards apply to the decisions the AI makes on your behalf, and whether your contract gives you any remedy if the answer to those questions changes after you have signed.

You cannot rely on the flag to tell you what the ship actually is. You have to look behind it.

The non-human corporation is not yet a practical reality in Australia. It is a signal, though, that the commercial and legal structures that underpin supply chains are under genuine pressure. The maritime sector spent decades discovering that a system designed to attract investment can become a system that conceals harm. Flags of AI convenience will follow the same arc: faster, at greater scale, and with fewer physical anchors for regulators to hold on to. Procurement's job has always been to understand what sits behind a supplier, not just what a supplier provides. In an era where that question may soon have no human answer, the function that asks it most rigorously will be the one that protects its organisation best.

Procurement's job has always been to understand what sits behind a supplier, not just what a supplier provides.

‍

References

[1] Let's Data Science, 'Milei Proposes Non-Human Corporations for AI Agents', 12 June 2026. https://letsdatascience.com/news/milei-proposes-non-human-corporations-for-ai-421b6c02

[2] Buenos Aires Herald, 'Milei's proposal to allow non-human corporations run by AI causes concern in Argentina', June 2026. https://buenosairesherald.com/business/tech/mileis-proposal-to-allow-non-human-corporations-run-by-ai-causes-concern-in-argentina

[3] The Legal Wire, 'Argentina floats non-human corporations run entirely by AI', June 2026. https://thelegalwire.ai/argentina-floats-non-human-corporations-run-entirely-by-ai/

[4] PYMNTS, 'Argentina Drafts Corporate Law That Requires No Human Boss', 4 June 2026. https://www.pymnts.com/news/artificial-intelligence/2026/argentina-drafts-corporate-law-that-requires-no-human-boss/

[5] Business Model Analyst, 'Argentina Wants AI to Own Companies, No Humans Needed', June 2026. https://businessmodelanalyst.com/argentina-ai-non-human-corporations/

[6] AI Weekly, 'Milei Proposes Non-Human Corporations for AI Agents', June 2026. https://aiweekly.co/alerts/milei-proposes-non-human-corporations-for-ai-agents

[7] Bird & Bird, AI Regulatory Horizon Tracker: Australia, updated June 2026. https://www.twobirds.com/en/capabilities/artificial-intelligence/ai-legal-services/ai-regulatory-horizon-tracker/australia

[8] IAPP, 'Australia unveils AI policy roadmap', February 2026. https://iapp.org/news/a/australia-unveils-ai-policy-roadmap

[9] APRA, 'APRA Letter to Industry on Artificial Intelligence (AI)', 30 April 2026. https://www.apra.gov.au/apra-letter-to-industry-on-artificial-intelligence-ai

[10] MinterEllison, 'Expert Guide: Legal Issues in Acquiring AI Companies', March 2026. https://www.minterellison.com/articles/expert-guide-legal-issues-in-acquiring-ai-companies

[11] Hall & Wilcox, 'AI procurement in Australia: trends, risks and what to expect in 2026', 2026. https://hallandwilcox.com.au/news/ai-procurement-in-australia-trends-risks-and-what-to-expect-in-2026/

[12] The Adaptavist Group, 'AI regulation in Australia 2026: Strategic guide for business leaders', February 2026. https://www.theadaptavistgroup.com/blog/ai-regulation-in-australia

[13] Viral Methods, 'Innovation: Argentina creates law for AI companies', June 2026. https://metodoviral.com/en/news/innovation-argentina-creates-law-for-ai-companies/

[14] ITF Global, 'Flags of Convenience', Supply Chain Guidance. https://www.itfglobal.org/en/supply-chain-guidance/risks-sector/maritime-shipping/structural-risks/flags-convenience

[15] SAFETY4SEA, 'ITF issues three new additions to Flags of Convenience list', December 2025. https://safety4sea.com/itf-issues-three-new-additions-to-flags-of-convenience-list/

[16] ITF Seafarers, 'Flags of Convenience'. https://www.itfseafarers.org/en/issues/flags-of-convenience

[17] Splash247, 'ITF says flags of convenience must shape up', April 2016. https://splash247.com/itf-says-flags-of-convenience-must-shape-up/

[18] Seafarers International Union, 'SIU, ITF Back Investigation into Flags of Convenience', July 2025. https://www.seafarers.org/seafarerslogs/2025/07/siu-itf-back-investigation-into-flags-of-convenience/

[19] Context News / Thomson Reuters Foundation, 'How shipping's flags of convenience endanger seafarers', August 2025. https://www.context.news/money-power-people/how-shippings-flags-of-convenience-endanger-seafarers

[20] Oxford Public International Law, 'Flags of Convenience', Oxford University Press. https://opil.ouplaw.com/display/10.1093/law:epil/9780199231690/law-9780199231690-e1167