Agents cross company lines. 7 trust designs.

For most of the last three years, AI governance has been an internal matter. A company decides which tools it allows, who approves them and how it checks them. The AI Governance Setup and the AI Trust Record do exactly that, and they remain the right first steps.

Agents change the question. A purchasing agent places orders with a supplier's sales agent. A customer's assistant asks a company's service agent to change a booking. A research agent pulls data from a partner's system. In each case, an AI system acting for one organisation meets an AI system acting for another. Neither side's internal approval tells the other side anything. The insight on the governance gap that agents open describes the problem in detail.

Seven designs

The trust infrastructure class collects ETHORITY's research and architecture work on that problem. All seven are shown today as designs, with their purpose, structure and open questions.

AI Trust Profiles connect an AI system, its intended use and the evidence relevant to that use. They extend the profile cards of the Good AI Prism from "what this product is" to "what this deployment may do".

Trust Passport explores how evidence could travel with a system while keeping its identity, scope and validity conditions intact, so that a statement made about a system in one context is not silently reused in another.

Trust Handshake explores how two systems could establish identity, evidence and permission before they interact, the AI equivalent of checking a counterparty before signing.

Agent Authorization & Trust Protocols make an agent's actor, scope, constraints and human intervention points explicit in a protocol design. The insight on agent authority and its limits shows why: an agent's effective reach is often larger than anyone approved.

Trust Network defines participation around distinct identities, evidence, obligations and admission conditions. Trusted Capability Marketplace designs discovery around capability identity, evidence and the permissions needed for use.

The AI Business Operating System connects intelligence, decisions, governed work and improvement in one architecture. ETHORITY is built as an AI-first organisation and runs its own research and delivery on this architecture; the SaaS edition is a design for others.

Why build on the Trust Chain

All seven designs extend one structure that is already in use: the Trust Chain. In every Trust Record, it links the outcome a system should deliver to its owner, data, AI component, human authority, evidence and review. The designs ask what happens when that chain has to be read by someone outside the company, or by another system.

The AI Act transparency duties under Article 50 have applied since 2 August 2026: people must be told when they interact with AI. Between organisations, the same principle points further. A system should be able to show who it acts for, what it may do and what evidence stands behind it.

Research first, product later

These designs are not sold as live services, and the pages say so plainly. Credentials, networks and marketplaces need identity infrastructure, security, legal agreements and operating support that must exist before anyone relies on them. Showing the architecture now lets clients, providers and researchers shape it while it is still a design.

What clients can use today is the foundation: a Trust Record decision document for one system, a governance model for the organisation and an estate inventory that knows which agents exist. Organisations whose agents already act across company boundaries are invited into the design discussions, and the insight on buying an AI agent sets out the questions to ask in the meantime.

What would make it real

Three conditions decide when a design becomes a product. Identity: participants must be able to prove who they are and who their agents act for. Evidence: statements about a system must carry their source, date and scope, the way registry facts already do. Authority: every agent action must map to a person or role that permitted it. The research tracks progress on each condition in public, so the move from architecture to product will be visible when it happens. Until then, the most useful thing an organisation can do is to write down, for each agent it runs, who it acts for, what it may do and what evidence stands behind it. That record is the seed every one of these designs grows from.