Attest Model

Create a provenance attestation for an AI model, linking it to training datasets and recording framework metadata.

Creates a cryptographically anchored attestation for an AI model. The attestation captures the model’s framework, training lineage, hyperparameters, and evaluation metrics. It is linked to the originating agent and anchored in the transparency log.

Model attestations provide an auditable record of what was trained, how it was trained, and what data was used — enabling compliance teams and downstream consumers to verify model provenance before deployment.

Authentication

X-API-Key · string · required

API key with models:write scope. Alternatively, pass a Bearer JWT token in the Authorization header.

X-Tenant-ID · string · required

Tenant identifier for multi-tenant isolation.

Request

agent_id · string · required

MAIP agent identifier that trained or owns this model.

name · string · required

Human-readable model name. Must be unique within the agent’s namespace per version.

version · string

Semantic version of the model (e.g. 2.1.0). Defaults to 1.0.0 if omitted.

framework · string

ML framework used to train the model. Accepted values: pytorch, tensorflow, onnx, custom.

hash · string

SHA-256 hex digest of the serialized model weights. Used for integrity verification at deployment time.

training_dataset_ids · string[]

Array of dataset attestation identifiers (maip-ds:ULID) used to train this model. Creates lineage edges automatically.

hyperparameters · object

Key-value map of hyperparameters used during training (e.g. learning_rate, batch_size, epochs).

metrics · object

Key-value map of evaluation metrics (e.g. accuracy, f1_score, auc_roc).

Response

id · string

Unique model attestation identifier in MAIP format (maip-model:ULID).

agent_id · string

The agent that attested this model.

name · string

Model name as provided in the request.

version · string

Resolved version of the model.

framework · string

ML framework identifier.

hash · string

SHA-256 hex digest stored for integrity verification.

training_dataset_ids · string[]

Linked training dataset attestation identifiers.

status · string

Attestation status. Always attested on creation.

receipt_id · string

Transparency-log receipt anchoring this attestation.

created_at · string

ISO 8601 timestamp of creation.