EMILIA is being developed as an open protocol for trust decisions and appeals. The protocol layer should become stronger through inspectability, conformance, ecosystem participation, and broader governance over time.
We believe the protocol layer and the commercial layer should be clearly legible. The protocol can remain open and portable while companies build products, services, and implementation support on top.
The protocol should remain inspectable, interoperable, and challengeable even as commercial products are built on top of it.
Our long-term direction is to support broader participation in governance, conformance expectations, and policy discussion as the ecosystem matures.
Dispute resolution is no longer purely operator-managed. High-confidence vouchers in the trust graph now vote on contested receipts. Voting weight is proportional to accumulated evidence — it cannot be purchased or injected. This makes the adjudication process Sybil-resistant by design and structurally harder to capture by any single operator.
The 48-hour procedural window before graph adjudication is enforced in code, not just policy. The dispute lifecycle — submission, operator response window, escalation to graph vote — is executed by the protocol itself. No manual override is needed; no human can short-circuit the window.
Every receipt now carries an attribution chain: Principal → Agent → Tool. This creates a verifiable record of which human authorized which agent action, executed through which tool. Accountability for agent behavior is not diffused — it traces back to a specific human delegation decision.
Delegation Authority extends this further: EMILIA now scores the quality of human delegation decisions, not just agent outcomes. Principals who consistently authorize well-scoped, low-risk delegations build positive reputation. Principals who authorize reckless or disputed actions accumulate negative signal. Human accountability for machine behavior becomes legible and contestable.
We are looking for technical reviewers, governance participants, and ecosystem partners who can strengthen correctness, adoption, and legitimacy.