“A significant security risk of using foreign-developed AI coding models is the potential for them to embed hidden vulnerabilities in generated code, which could be exploited by that government's intelligence agencies.”