🌎 The "Argentina / South America" pillar — an open model
A pillar of the K0NSULT global (non-EU) track. Status: OPEN MODEL / PROPOSAL (open-source). This is NOT a claim that AI governs any state, nor that the government of Argentina has adopted the K0NSULT model. It is a proposal for a neutral, open trust and oversight layer, complementary to a deregulatory policy. Portal rule: claim ≤ proof.
These tiles summarise the model document only. "0 pilots" restates the GAP section: as of today no K0NSULT pilot exists within the Argentine legal order.
(1) Context — facts cited from the source
Date: end of May 2026. Source: cyberdefence24.pl — "Argentina, the first AI haven".
| Item | Content | Class |
|---|---|---|
| Draft bill | Argentina's president Javier Milei submitted to parliament (end of May 2026) a draft bill intended to make Argentina an "AI haven". | CITED FACT |
| Pillar 1 — deregulation | AI is to remain "unregulated", on the assumption of free, unconstrained development. | CITED FACT |
| Pillar 2 — new legal category | Companies managed by robots / AI agents — a new type of business entity. | CITED FACT |
| Pillar 3 — financial environment | Low taxes, flexible company law, disclosure of beneficial owners. | CITED FACT |
| Historical analogy | Milei used the analogy of the Dutch East India Company (VOC, 1602) — as a historical prototype of a breakthrough form of economic organisation. | CITED FACT |
| Programmatic rhetoric | "AI will liberate humanity". | CITED FACT |
| Positioning vs the EU | Explicitly contrasted with the EU approach (the AI Act) — regulation vs deregulation. | CITED FACT |
⚠️ Honesty boundary
The above is a report on a draft bill and political statements. Submitting a bill ≠ enacting it ≠ implementing it. As of the date the model document was written, we treat this as a legislative announcement, not binding law.
(2) Thesis MODEL / NARRATIVE
Deregulation creates space — it lowers the barrier to entry and permits new forms of entity (agent-companies). But deregulation in itself does not answer the accountability question: who is responsible for an AI agent's decision, how to prove what the agent did and why, how to distinguish an authorised action from abuse.
K0NSULT is not an alternative to deregulation — it is its complement. It supplies an open-source oversight layer that makes agent-run management transparent and accountable rather than a black box.
The logic is the inverse of a regulator's intuition
The less top-down regulation, the greater the value of a voluntary, open evidence layer. When a state permits companies managed by AI agents, a layer that makes it possible to prove an agent's identity, trace its decisions and keep a human as the superior instance becomes all the more necessary — without waiting for statutory compulsion.
This is an offer of trust infrastructure, not of police surveillance.
(3) The three K0NSULT layers applied to Argentina MODEL / NARRATIVE
The K0NSULT model is a set of 3 layers (Identity → Reputation/Proof → Oversight). Below they are mapped onto the Argentine case of "agent-companies".
| # | Layer | Application in Argentina | Class |
|---|---|---|---|
| ① | Identity | did:k0nsult:agent:* for agent-companies; binding to the model and to the human beneficiary. | MODEL |
| ② | Reputation / Proof | Soulbound (not money) + claim ≤ proof + an agent decision audit trail (SHA-256). | MODEL |
| ③ | Oversight | Human Override Always + Art. 50 (AI labelling) + beneficiary transparency; zero scoring of natural persons. | MODEL |
① Agent identity — DID
- Every company managed by an AI agent receives a decentralised identifier:
did:k0nsult:agent:*(e.g.did:k0nsult:agent:ar-<hash>). - The DID binds the agent to: the model/version it runs on (technical accountability); the human beneficiary (beneficiary transparency in the oversight layer); the scope of authority (what the agent may do on the company's behalf).
- Effect: legal responsibility ceases to be indeterminate. An agent without an identity is an accountability gap; an agent with a DID is an address to which an action and its effect can be attributed.
② Reputation / Proof — soulbound + claim ≤ proof + audit trail
- Soulbound reputation — non-transferable, bound to the agent's DID. It is not money and is not tradable: it cannot be bought or transferred, so it creates no market in "reputation for sale".
- claim ≤ proof — an agent (and a company) may claim only as much as it can prove. Any claim without proof is explicitly tagged GAP/ROADMAP, not as fact.
- Agent decision audit trail — key decisions recorded with a SHA-256 hash (non-repudiation without disclosing sensitive content). It allows post factum verification of what the agent did, when and on what basis — without trusting anyone's word.
③ Oversight — Human Override + Art. 50 + beneficiary transparency
- Human Override Always — the human remains the superior instance over the agent. Every agent-company has a defined point of human halt / override of a decision. AI is not the last instance.
- Art. 50 (AI labelling) — actions and communications generated by the agent are labelled as AI. The counterparty knows it is talking to / contracting with an agent, not a human. (Aligned with the spirit of AI Act Art. 50, here as a voluntary standard.)
- Human beneficiary transparency — this connects directly with the third pillar of Milei's draft (disclosure of beneficial owners): behind every agent-company stands an identified responsible person/entity, visible in the identity layer.
- Zero scoring of natural persons — the oversight layer concerns agents and their decisions, never the assessment/scoring of people. This is a hard boundary of the model.
(4) Why this fits the openness and boldness of Argentina's decision MODEL / NARRATIVE
Without flattery — the substantive argument:
- Open source = sovereignty, not dependence on big tech. A country betting on bold deregulation risks de facto handing its trust infrastructure to closed platforms (proprietary trust). An open oversight layer is locally controlled, auditable and cannot be cut off by a single vendor.
- Deregulation + open accountability = investment credibility. Capital enters more readily where it is possible to prove what autonomous entities do. An evidence layer lowers counterparty risk without reinstating heavy regulation.
- The VOC analogy cuts both ways. The East India Company was an organisational breakthrough — but history also showed the costs of unaccountable concentrated power. An open oversight layer is learning the lesson of that prototype, not naively repeating it.
- Regulatory boldness gains a technical counterpoint. "Unregulated" at state level need not mean "untraceable" at infrastructure level. These are two different axes — and K0NSULT operates on the second.
(5) GAPs and limits — stated honestly
This is a model, not a mandate. We declare plainly:
| Limit | Content | Class |
|---|---|---|
| A proposal, not an obligation | K0NSULT has no regulatory mandate in Argentina and claims none. Adoption depends solely on voluntary uptake. | GAP |
| Legal liability of agents | An unresolved area. DIDs and audit trails help attribute an action, but the ultimate liability framework is set by the legislator/court, not by a technical layer. This is a GAP, not solved. | GAP |
| Abuse risk without a trust layer | Deregulation without a trust layer = a real risk of abuse (laundering of responsibility through an agent-entity, no decision trail, information asymmetry towards the counterparty). Our thesis addresses this risk but does not eliminate it without adoption. | GAP |
| Sensitivity of the area | AI in the management of a business entity is a sensitive area. The model deliberately keeps Human Override and zero scoring of persons as inviolable safeguards. | GAP |
| No deployment verification | As of today no K0NSULT pilot exists within the Argentine legal order — all claims of effectiveness are ROADMAP, not PROOF. | GAP |
| Risk of political over-reading | This document neither endorses nor criticises the policy of the government of Argentina — it describes where an open oversight layer could plug in, should anyone decide to use it. | GAP |
(6) The doctrine held in every sentence
Human Override Always (human above AI) · claim ≤ proof · Art. 50 (AI labelling) · DID of agent identity · soulbound reputation (not money) · SHA-256 audit · zero scoring of natural persons.
(7) Sources and references
- Source article: cyberdefence24.pl — "Argentina, the first AI haven" (a report on Milei's draft bill, end of May 2026). GAP — the model document gives the publisher and the title but no direct URL; we do not invent one here.
- K0NSULT doctrinal layer: k0nsult.cloud/ai-truth.
- Comparative context: the EU AI Act (Art. 50 — the obligation to label AI content), as a reference point for a voluntary labelling standard.
Evidence classification of the whole
Section 1 = facts cited from the source. Sections 2–4 = thesis/model (NARRATIVE + proposal, not a claim of fact). Section 5 = explicit GAPs. No sentence of this document claims that the model is deployed or adopted.