AI-generated code contains more bugs and errors than human output (www.techradar.com)

A recent report from CodeRabbit reveals that AI-generated code is more error-prone than human-written code, with an average of 10.83 issues per AI-generated pull request compared to 6.45 for human-generated requests. These discrepancies include 1.7 times more overall issues and a higher prevalence of critical vulnerabilities, such as improper password handling and insecure object references. While AI tools can enhance coding efficiency, they introduce significant bugs that can complicate the review process, raising concerns about their reliability in software development. Despite these drawbacks, the study highlights the evolving role of human developers, who may increasingly shift toward managing and reviewing AI-generated outputs rather than writing code from scratch. AI tools improve spelling and testability, which suggests a complementary relationship where AI accelerates initial development while humans oversee quality and security. As these AI models, like OpenAI's GPT family, continue to improve, the balance between productivity and code integrity will be crucial for organizations looking to leverage AI advancements in their programming practices.