"AI-generated code and human-written code exhibit different defect patterns, complexity profiles and security vulnerabilities in large-scale comparisons."
Cotroneo and colleagues compared more than 500,000 Python and Java code samples and identified distinct quality and vulnerability patterns in human-written and AI-generated code. The study supports the chapter's concern that machine-generated artifacts can present unfamiliar engineering failure modes.
View Source