I don't know if I missed something, but this CVE isn't that major as it was suggested to be? For one it had to originate from app.opencode.com and even if it didn't most (good) browsers block websites from probing localhost. Yes it is still a pretty bad CVE, but not as critical as some might suggest.

I have a hypothesis why issues like that are so widespread. That AI infrastructure is mostly developed by large companies; their business model is selling software as a service at scale. Hence containers, micro-services, TCP/IP in between. That approach is reasonable for data centres because these made of multiple servers i.e. need networking, and they have private virtual networks just to connect servers so the security consequences aren’t too bad.

If they were designing these infrastructure pieces primarily for consumer use, they would have used named pipes, Unix domain sockets, or some other local-only IPC method instead of TCP/IP.

Great write up.

These local agents that you spawn and give access to your drive are kind of insane to me.

It's at the level of

     /bin/bash -c "$(curl -fsSL https://somescriptofftheinternet

which you cannot inspect, and may be well different every time you interact with it!

As per usual, being at the forefront of the tech world is leaving behind privacy and security in the dust... until something bad happens.

> a RCE vulnerability is the type of thing that nation state actors in Russia and North Korea dream of

Does this mean other state actors are beyond needs of RCE vulns as their tools belt and North Korea and Russia lagging behind? Some other interpretation from security-involved practitioners here - like, I don't know - we already have Pegasus, phew on OpenCode RCE?

The one thing here confusing to me is the past tense used throughout. This CVE seems presented as both past and present, yet the present evidence isn't... Presented.

Wild that people don't run these kind of AI agent tools in isolated containers. They seem crazy dangerous even without CVEs.