Devs don't trust AI code - but many say they still don't check it anyways (www.techradar.com)

A recent survey by Sonar reveals that a staggering 96% of developers lack full trust in AI-generated code, yet nearly half do not consistently check for errors in that code. The survey highlights that 42% of developers’ code is AI-generated, a figure projected to rise to 65% by 2027. Despite the significant adoption of AI tools like GitHub Copilot and ChatGPT—used by 75% and 74% of developers respectively—more than one-third opt to use personal accounts rather than company-sanctioned ones, raising concerns about data exposure and security vulnerabilities. This lack of oversight is alarming, especially as AI-generated code reportedly poses 1.7 times more issues than human-written code. While AI can significantly enhance coding efficiency, the findings underscore the critical need for developers to verify AI outputs properly to prevent the introduction of bugs and vulnerabilities into production environments. As AI tools become increasingly integrated into the development process, ensuring the trustworthiness and verifiability of AI-generated code is paramount for the integrity and security of software development.