Which AI models can be evaluated using the A1 Security Assessment?

The A1 Security Assessment module evaluates the security, governance, and risk management of artificial intelligence models. HITRUST specifies coverage for widely used model types, including:

Predictive models, which forecast outcomes based on historical data (e.g., fraud detection, patient risk scoring).

Generative models, which create new data outputs (e.g., AI image or text generators).

Rule-based models, which use defined logic for decision-making.

The goal of the A1 assessment is to ensure that these AI models are developed, implemented, and monitored securely, with appropriate safeguards around data integrity, bias management, and model explainability. Options like Hodgkin-Huxley (a neuroscience model) and Back Propagation (a training algorithm) are not types of AI models scoped by the A1 assessment. Instead, the A1 factor focuses on applied model categories used in operational environments.