AI/ML needs the same type of accountability and transparency that we demand for the open-source or proprietary software we create and distribute. AI models are now being used for various use cases but knowing and understanding how the model was trained is paramount as false outputs could lead to compliance, security, and reputational risk. For example, a bank integrates with an AI model to help them run loan risk assessments. Depending on the training data and software supply chain, the AI model may deny loan approval due to race, sex, national origin, and other forms of illegal discrimination. This would lead to a compliance risk as the bank is not meeting the regulations, a security risk as the AI model was tainted somehow, and a reputational risk as the bank would lose customers based on misdeeds. Therefore, it is important to understand the provenance of the data the models were trained on, and if an incident does occur, we can quickly determine the source and where else this tainted data might have been used. Since models are not inspectable by themselves, the only way to ensure they behave as intended by the trainer is to create tamper-proof evidence of the training process and use that for any model behavior analysis.