供應商不透明阻礙風險評估

Risk Description

AI model vendors refuse to disclose training data sources, architecture details, known limitations, and safety test results citing trade secrets. Deploying organizations cannot conduct adequate risk assessments without this critical information, effectively "deploying blind." This vendor opacity is particularly severe under AI-as-a-Service models, as organizations rely entirely on limited vendor-provided information for deployment decisions.