PRAMETRA
Menu

Security

Secure by design. Verifiable by evidence.

PRAMETRA is designed around explicit authority, bounded capability, examinable evidence and deterministic final-action boundaries.

This page describes design principles and intended control boundaries. It is not a claim of certification, external audit outcome, production deployment status or security maturity.

Public principleCapability does not create authority.

Security principles

Authority cannot depend on a model behaving correctly.

Generative reasoning may inform a decision. It must not grant itself identity, permission, approval or the authority to execute a consequential action.

This page describes design principles and intended control boundaries. It is not a claim of certification, external audit outcome, production deployment status or security maturity.

  • 01

    Identity and authority

    Identity and delegated authority are designed to remain explicit and outside generative self-assertion.

  • 02

    Capability is not authority

    Technical ability to propose or perform an action does not establish permission to do so.

  • 03

    Least privilege

    PRAMETRA's principles require access and action rights to be limited to the minimum necessary scope.

  • 04

    Controlled action

    Action should progress only through explicit, bounded and reviewable control points.

  • 05

    Provenance

    Material sources, transformations and decision inputs should remain attributable.

  • 06

    Evidence integrity

    Evidence, inference and contradiction should remain distinguishable throughout the decision record.

  • 07

    Explicit approval

    Approval requirements should reflect authority, policy and potential consequence.

  • 08

    Deterministic final action

    Final-action boundaries are designed to remain deterministic even when reasoning engines are probabilistic.

  • 09

    Recovery and reconstruction

    Decision state and material evidence should support safe recovery, review and reconstruction.

Decision boundary

Reasoning may vary. Control must remain explicit.

  1. ReasoningReplaceable models and agents

    Generate hypotheses, interpretations and recommendations within a bounded context.

    01
  2. Deterministic boundaryIdentity, authority, policy and approval

    Establish who may decide, what constraints apply and whether action may progress.

    02
  3. AccountabilityHuman-governed action and verification

    Keep accountable approval, observed outcomes and later review connected to the decision.

    03
Explore the Platform