Enhancing MLOps Security: Preventing Unauthorized Model Publishing on SAS Viya

Published on May 22, 2026, on SAS Communities, an article outlines crucial strategies for preventing unauthorized publishing of machine learning models within the SAS Viya environment, a critical aspect of secure MLOps. The piece highlights the intricate collaboration required among data scientists, MLOps engineers, and risk management teams to maintain model integrity and compliance. The article introduces three primary methods to enhance security. Firstly, it details the integration between SAS Model Manager, used by data scientists and MLOps engineers for model deployment and monitoring, and SAS Model Risk Management (MRM), utilized by risk analysts. By linking models within MRM, a crucial control is established: a model cannot be published from Model Manager until it has received the necessary risk sign-off and approval. This linkage ensures that models undergo rigorous auditing processes before entering production, fostering efficient communication and shared information between development and risk teams. Secondly, the article discusses the implementation of granular user permissions. Administrators can configure prohibit rules for specific users or groups, preventing them from publishing models even if they can view them within the repository. This method provides a strong access control layer, ensuring that only authorized personnel can initiate model deployments. Lastly, the use of custom workflows, managed via SAS Workflow Manager (included with SAS Model Manager), is presented as a powerful tool. These workflows can be designed to represent and enforce standardized business processes for model validation and approval. By integrating these workflows, organizations can create a structured, auditable path for model promotion, further safeguarding against unauthorized publishing and ensuring adherence to internal governance policies and external regulations. These combined approaches significantly strengthen the security posture of MLOps pipelines on SAS Viya.