An Al-Driven Cybersecurity Framework for Prompt Security, Repository Vulnerability Detection, and Static Code Analysis

Abstract

The rapid adoption of artificial intelligence (AI), large language models (LLMs), and AI-assisted development tools has introduced new security challenges that traditional application security mechanisms do not fully address. Threats such as prompt injection, sensitive information disclosure, inse-cure output handling, excessive agency, vulnerable dependencies, hard-coded secrets, and AI supply-chain attacks can affect both AI components and conventional software infrastructure. This paper presents an AI-Powered Security Intelligence Platform for Developers that integrates source-code analysis, dependency security analysis, AI threat detection, threat intelligence, risk pri-oritization, and security recommendations into a unified frame-work. The proposed architecture combines conventional security analysis with AI-assisted reasoning to identify and contextualize security weaknesses across the software development lifecycle. Detected threats are mapped to established security knowledge bases including OWASP GenAI LLM Top 10, NIST Artificial Intelligence Risk Management Framework, MITRE ATLAS, CWE, and CVE/NVD. The platform generates contextual security intelligence containing vulnerability severity, potential impact, attack vectors, related security classifications, and recommended remediation. The proposed framework provides a foundation for integrating AI security analysis with developer-oriented vulnerability management and continuous security monitoring.