Method

Method and limits

This page describes the method behind the developing Sovrentic assurance scheme and the claims we do not make while its operating model is in development.

The question we answer

Why may this configuration of this agent be used for this task, and what changes when its configuration or evidence changes?

That question forces clarity on identity, scope, demonstrated capability, evidence and the human decision that binds them together. It is the organising idea behind a bounded Sovrentic assurance record.

The five connected steps

  1. Identity

    Who or what is acting, and on whose mandate. An agent is not just a model; it is a configured runtime with permissions, tools and an accountable owner.

  2. Configuration

    The boundaries, policies and restrictions the agent may operate within. Change the configuration and the assurance position must be reassessed.

  3. Assessment

    What the configuration was demonstrated to do, under which conditions, with what limitations. A claim without scope is not evidence.

  4. Evidence

    A traceable, time-bound record that links claims to identifiable provenance. Evidence must be inspectable by authorised reviewers.

  5. Decision

    The explicit point where a human owner permits, limits, reviews or suspends use. The authority is retained, not delegated away.

What we do not claim

  • Sovrentic is not operating a live assessment service today and is not an accredited conformity-assessment body.
  • We do not claim independent validation, customer results or production availability.
  • No customer deployment, operational assessment service or live issuance pipeline is implied on this site.
  • A record or status shown in product concepts is illustrative, not a guarantee of agent safety or universal compliance.
  • The final decision and accountability always rest with the human owner.

The assurance lifecycle

A Sovrentic assurance record is not a permanent badge. It is a bounded claim about a defined agentic resource, assessed against an explicit profile and revisited as its configuration, evidence or operating context changes.

  1. 01RegisterIdentify the resource, its version and its accountable owner.
  2. 02AssessTest the declared scope against the applicable profile.
  3. 03AttestRecord a bounded claim with its conditions and limitations.
  4. 04VerifyLet a relying party check status, scope and validity.
  5. 05MonitorWatch for material change in configuration, evidence or context.
  6. 06Reassess or revokeRevisit the claim when the conditions no longer hold.

Optional public anchoring may support integrity, provenance and transparent record history. It does not by itself prove behavioural truth, runtime identity or current non-revocation.

Development stage

The Sovrentic assurance scheme is currently in development. Founding design partners and qualified assurance organisations are invited to shape the first operating profiles for defined capabilities, evidence and bounded decisions.