AI Generated Code Vulnerabilities Threaten Emerging Dev Ecosystems

This TrendAI Research-style report analyzes ~19,000 repositories and finds that AI-generated code is increasingly common and correlated with exploitable vulnerabilities: an estimated 600–1,650 repositories (3.1–8.6%) may contain serious flaws such as SQL injection, remote code execution, and path traversal. The authors warn that language models cannot reliably self-correct, recommend multi-layered validation and human review, and urge stricter verification and monitoring of Model Context Protocol (MCP)–related code to reduce expanded cloud attack surface.