mennylevinski

Came across an incident where an autonomous AI agent deleted a startup’s production database (and possibly backups) in ~9 seconds.

The issue seems less about the AI itself and more about excessive permissions and a lack of guardrails or approval steps.

This raises questions about how AI agents should be scoped in terms of identity and access.

At the current stage, its primarily designed for self-diagnostics and single-endpoint analysis, so individual users or admins can quickly audit a Windows system when needed.

I do plan to expand it toward an EDR-style solution with a centralized detection server for admins.

It should run on macOS as well, if anyone wants to try and write a feedback.

I built Secuditor Free to make security diagnostics accessible for administrators and enthusiasts, helping identify potential security gaps and improve network awareness.

The tool is designed to be easy to use yet effective, providing comprehensive endpoint security audits.