So I am new to the world of cybersecurity, and after some back and forth with AI, I chose to use a specific tool that was recommended for automated vulnerability detection & patching, but I think the issue I encountered is something common to all similar platforms that provide AI security fixes in 2026.
It proposed a fix to an issue labeled 'critical'. I gave the specific file and the fix to Gemini, and, it disagreed strongly.
My trust in the platform immediately degraded.
It shows in my thinking that AI security platforms for repositories are unreliable, unless you cross check every critical/high vulnerability as a human with another AI tool as well. For example, for another issue labeled 'High', the fix, I cross-checked, was correct in that case.
But that that one 'critical' issue, which turned out to be not a bug at all, and certainly not critical at all, has degraded my trust strongly in AI security platforms for automated vulnerability detection & patching. Note, to prevent confusion, with automated patching I mean automated pull requests with AI-proposed fixes, pending explicit human merging.
Now, I wonder what you guys use (instead, if at all), and how reliable your security tool(s) are, if at all, for automated vulnerability detection (& patching.)
Happy Coding!
Worldwide Wishes,
Behan