What is deterministic identity for AI?

Deterministic identity for AI is the assignment of a stable, verifiable identity to AI run declarations derived from canonical representation. Identical input bytes always produce identical identity. It is not related to deterministic execution or predictable AI outputs.

Is deterministic identity the same as deterministic AI?

No. Deterministic AI controls execution behavior to produce predictable outputs. Deterministic identity assigns a cryptographic identity to declarations independent of execution. These are separate concerns.

How is deterministic identity for AI computed?

Input bytes are parsed, serialized into canonical JSON, and hashed using SHA-256. The resulting digest is the identity. All compliant implementations produce identical identity for identical input bytes.

Deterministic Identity for AI

Deterministic identity for AI refers to the assignment of a stable, verifiable identity to AI run declarations derived from canonical representation — not execution behavior.

Deterministic identity for AI does not refer to deterministic behavior or predictable outputs.

What Deterministic Identity for AI Is Not

Not deterministic outputs
Not execution control
Not workflow logic
Not model reproducibility

These are execution-level concerns. Deterministic identity operates at the declaration level, before execution occurs.

Definition

Deterministic identity for AI run declarations is the cryptographic identity derived from canonical JSON bytes representing declared execution. Identical input bytes always produce identical canonical JSON and identical SHA-256 identity across all compliant implementations.

This identity is stable, implementation-independent, and verifiable without requiring trust in execution.

Identity Model

The deterministic identity model follows a fixed path:

Input bytes are parsed under protocol rules
The declaration is serialized into canonical JSON
Canonical JSON bytes are hashed using SHA-256
The resulting digest is the identity

Identity is independent of execution. Two systems processing the same valid declaration always converge on the same identity.

Why Deterministic Identity for AI Matters

Auditability: every declaration has a verifiable, tamper-evident identity
Cross-system verification: independent systems confirm identity equality without coordination
Portability: identity travels with the declaration, not with the system that produced it
Tamper detection: any modification to the declaration changes the identity

Without deterministic identity, AI declarations cannot be reliably referenced, compared, or audited across environments.

How SCL Implements This Model

SCL is the deterministic, cryptographic identity protocol for AI run declarations — see the canonical definition. It implements the deterministic identity model described above through a frozen specification with strict canonicalization and hashing rules.

See the protocol definition and processing pipeline
The SCL:V1 specification defines the complete rules
The reference engine verifies compliance
See how identity applies to declarations for application details

Deterministic Identity vs Deterministic AI

Deterministic identity and deterministic AI are not the same concept. See the full comparison of deterministic AI vs AI identity protocol.

Deterministic identity for AI is a representation-level property, not an execution-level property.

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