When AI-built apps get risky: data from 3,911 repos (pathtoship.com)

🤖 AI Summary
A recent analysis of 3,911 repositories across nine AI application platforms revealed that while security flaws are prevalent in AI-built apps, the majority of these vulnerabilities exist in dormant projects that lack users. Initially reported at 98%, the actual percentage of apps with security flaws was revised down to 78.3%. Notably, 66.3% of the apps had no activity or user engagement, suggesting that many flaws would not pose a significant risk since they are not actively used. This insight challenges the perception that AI applications are broadly insecure and underscores the distinction between public GitHub repositories and deployed applications. Furthermore, the study highlighted a consistent trend: the likelihood of an app containing at least one critical security finding increases with the app's size. While apps with less than 1,000 lines of code had only a 2.9% chance of containing a critical flaw, that number jumped to nearly 60% for apps exceeding 40,000 lines. This indicates that as applications scale, the exposure to potential vulnerabilities grows simply due to increased code complexity. Developers are encouraged to prioritize security checks, especially for larger applications, where the risk of overlooked vulnerabilities becomes more pronounced.
Loading comments...
loading comments...