Abstract: Disclosed are techniques for security architecture generation using artificial intelligence. A computing system receives an architecture artifact and applies a computer vision model to extract an architectural component set and interconnection information, including icons and directionality information. The system generates a node collection based on the extracted components and performs vulnerability analysis by transforming nodes into vector embeddings and performing policy embedding search operations using vector similarity matching against a vector database storing policy data. The system generates a vulnerability artifact comprising a component set, a vulnerability set, and a policy item set, and applies an artificial intelligence model to generate a remediation recommendation specifying architectural changes.