Systems and methods for provable provenance for artificial intelligence model assessments

A system for providing provable provenance for an assessment, the system comprising:

• network circuitry, and
• control circuitry configured to:
• receive at least one of a dataset and a trained machine learning model over the network circuitry;
• select an assessment framework based on the at least one of the dataset and the trained machine learning model;
• generate assessment results by applying the assessment framework to the at least one of the dataset and the trained machine learning model;
• generate a hash value of the at least one of the dataset and the trained machine learning model;
• generate combined results comprising the hash value and the assessment results;
• encrypt the combined results using a private key to generate signed results;
• and
• transmit, over the network circuitry, the signed results and a certificate comprising a public key corresponding to the private key;
• and
• a client device configured to:
• decrypt the signed results using the public key;
• receive the at least one of the dataset and the trained machine learning model;
• identify a decrypted hash value based on the decrypted signed results;
• generate a hash value of the received at least one of the dataset and the trained machine learning model;
• and
• determine that the, proof of the provenance fails when at least one of:
• decryption, by the client device the control circuitry, of the signed results cannot be performed using the public key,
• a decrypted hash value does not match the hash value,
• the generated hash value of the received at least one of the dataset and the trained machine learning model does not match the decrypted hash value.