Verifiable AI Infrastructure

These terms describe approaches for increasing trust in AI systems by adding proofs, attestations, or isolated execution.

Trusted Execution Environment

A trusted execution environment (TEE) is a hardware-backed isolated environment that can protect code and data during execution and produce attestations about what ran.

ZKML

ZKML refers to techniques that generate cryptographic proofs about a machine learning model's computation, aiming to verify claims without revealing all inputs or internals.

Model Attestation

Model attestation is producing evidence about the model version and execution environment used for a result, helping downstream systems reason about trust and provenance.

Proof of Inference

A proof of inference is evidence that a particular model inference computation was performed as claimed, which can be used to improve trust in off-chain or distributed AI services.