When AI Gets It Wrong: The Insecure Defaults Lurking in Your Code

AI coding assistants accelerate development but also introduce security vulnerabilities by learning from insecure public code. Common AI-generated flaws include SQL injection via string concatenation, weak cryptography (MD5/SHA-1), hardcoded secrets, and vulnerable package recommendations. Developers should watch for four 'AI code smells': hardcoded credentials, hallucinated logic with unsafe defaults, outdated package suggestions, and high blast-radius refactors. The recommended fix is integrating real-time security scanning directly into the IDE rather than relying on post-commit scans, with Checkmarx Developer Assist presented as a solution that catches and auto-fixes AI-generated vulnerabilities inline.