Do AI assistants make coding easier or riskier? That seems to be the question at the heart of a recent study by Stanford University computer scientists, which reveals that developers using AI-powered assistants often produce buggier code.

The study, titled “Do Users Write More Insecure Code with AI Assistants?”, examined developers’ use of AI coding assistants, focusing on the controversial GitHub Copilot. The results of the study were revealing: participants with access to an AI assistant often produced more security vulnerabilities than those without.

So what can developers do to minimize bugs and make sure code is secure? First, developers should not blindly trust AI assistants that they are using. AI assistants are only as good as the data used to train them, and the accuracy and quality of their predictions depend on the data available. Additionally, developers should be aware of their own fallibility and take their own coding experience into account when making decisions about fixes.

Developers should also always test their code both manually and with automated testing tools. This is especially important when using AI assistants and automated code testing – no AI assistant is perfect, and developers should be prepared for the possibility of unexpected and unintended errors or bugs.

Lastly, developers should understand that even with AI and automated code testing, there is no substitute for the domain knowledge, experience, and expertise of a human programmer. AI assistants and automated tools can supplement the knowledge of a developer, but they can’t replace it.

In conclusion, while AI assistants may make coding more efficient, they should not be viewed as a replacement for human expertise. As the Stanford study shows, developers using AI-powered assistants often produce more bugs, making it even more important to stay vigilant and test code thoroughly.