Skip to main content

Data Schemas

Class Diagram​

Claims & Attestations​

Below are a few examples of claims and attestations which can be made on an AI system

Data​

  • Unwanted Bias:
    The presence of biases in the training data that can lead to skewed results or unfair outcomes.

  • Hallucinations:
    Instances where the system generates outputs that are factually incorrect or misleading due to inaccuracies in the data.

  • Errors in Generated Data:
    Refers to inaccuracies or mistakes in the data produced by the system during its operation or training.

  • Data Poisoning:
    The risk that adversarial inputs can corrupt the training dataset, potentially leading to malicious outcomes.

  • Data Pollution:
    The introduction of unwanted or low-quality data that degrades the quality and performance of the trained model.

Systems​

  • Cybersecurity Flaws:
    Vulnerabilities in the system architecture that could be exploited by malicious actors to compromise security.

  • Implementation Flaws:
    Issues arising from incorrect implementation of the system, which can lead to performance problems or security vulnerabilities.

  • Compliance Gaps:
    Potential areas where the system may not meet regulatory or industry standards, impacting trust and legal standing.