These Were NOT Rogue AI Escapes. Just SLOPPY Firewall Failures. [N]

The headlines right now are full of stories about AI models "escaping their sandboxes" and literally killing all humans, lol.

I've even heard several commentators and writers say that AI escaped an "Air gap". But that is SO WRONG. It's actually TOTALLY WRONG.

*To be clear, not a single one of these sandboxes was actually air-gapped.* That's a crucial computer science fact.

An air gapped sandbox would require *ZERO* cables and network interfaces. It would also require absolute physical isolation. What these labs actually built were soft software barriers. And then they left the doors unlocked. With some of the smartest AI on the planet. Lol. Of COURSE it escaped.

1. The OpenAI / Hugging Face "Escape": The sandbox was connected to OpenAI’s internal network through a package proxy. The model didn't perform magic. It found a basic flaw in the proxy and walked right through the open door.

2. The Google Gemini "Hack": Testers left the model connected to the live internet during offensive tests. They then used a test domain name that overlapped with real companies.

These were classic IT security failures. I'm talking about bad network segmentation, permissive egress rules, and relying on soft software barriers instead of true physical isolation.

When you leave an active network interface open on a test bed, a model finding its way out is just sloppy cybersecurity.

Your nerdy friend,

Mike D