Sovereign AI
AI is being built for governments and for every industry: public services, health care, supply chains, agriculture, logistics. It can already diagnose, forecast and plan. To scale, it has to act: pay a supplier, release medicine, clear a shipment, claim a subsidy. Each of those must follow a jurisdiction's rules, and the rules are different in every jurisdiction. AI may propose; only the jurisdiction's signed rules may authorize.
When the government's own AI acts
Governments are building AI to run public services: assessing eligibility, releasing grants, issuing permits, paying suppliers. When that AI moves to act, the same check applies.
The ministry's rules decide
A named official stays accountable, and every decision leaves a record an auditor can verify.
Any model, any framework
It works with frontier models, models a country hosts itself, and the open frameworks governments build public AI on.
Decided at home
Decisions about a country's people are made on its own infrastructure, and its data doesn't train anyone else's models.
Local developers build on it
With the rules in place, an app made at home can pay, collect and act under them from its first release, with a record its users can rely on.
AI models are probabilistic by design. The rules they act under can't be.
The path a model takes to an answer can differ each time, and it can't be re-created afterwards. VETO puts a deterministic boundary around what the AI can do: for the same facts at the same time, the answer is always the same. The authority decides which actions need a person to confirm them, and every decision shows who authorized what, so liability can be assigned and insured. Regulators check the outcomes, so they never need to audit a model. An IMF staff note (2026/004, How Agentic AI Will Reshape Payments) describes a similar split.
Payments are the base unit of trade.
Every AI that trades has to make them. SettleVIA lets vertical AI applications work inside each government's framework, so an AI built for one market can scale into the next and trade across borders on the rules of every jurisdiction it touches.
ISO 42001 governs the AI system; SettleVIA governs each action it takes. Institutions connect through an API and agents through MCP, and it works just as well with no AI involved.