What Deterministic Identity Covers in AI Run Declarations

For the definition of deterministic identity for AI, see the primary definition page.

Understanding what is SCL and what deterministic identity covers is essential to using the protocol correctly.

Deterministic identity for AI run declarations applies the deterministic identity model to structured declarations defined under the SCL protocol. This page defines what deterministic identity covers and does not cover in AI run declarations.

Identity Assignment Model

SCL assigns identity to AI run declarations through a fixed process. Each valid declaration receives exactly one identity derived from its canonical representation.

Two independent systems processing the same valid declaration always produce the same identity. The identity represents what was declared, not what was executed.

What the Identity Covers

Identity applies to the complete declaration as defined by the SCL:V1 grammar:

The identity is derived from the canonical JSON serialization of these components, hashed using SHA-256.

What the Identity Does Not Cover

The identity does not cover execution behavior, runtime outcomes, model outputs, or side effects. It confirms only that the declared content is identical at the byte level.

This makes identity useful for audit, cross-system referencing, and verification without requiring trust in execution.

Verification

All compliant implementations produce identical canonical JSON bytes and identical SHA-256 identity for identical input bytes. Any deviation constitutes protocol failure.

The complete rules are in the SCL:V1 specification. The reference engine verifies compliance.

Scope

SCL defines identity only. The canonical definition states the protocol premise. Scope boundaries are defined in What SCL Is Not.

Canonical Definition: SCL is the deterministic, cryptographic identity protocol for AI run declarations.