The most effective way to protect proprietary AI model components--architecture, weights, hyperparameters--is through strict access controls and encryption (option D).
AAIA identifies model theft and reverse engineering as high-severity security risks, especially for organizations whose competitive advantage relies on intellectual property embedded in AI models.
Proper controls include:
Role-based access control (RBAC)
Encryption at rest and in transit
Model-level access segregation
Hardware security modules (HSMs)
Zero-trust access principles