News for hackers. Stuff that matters.
Re:Claude Code 2.0 by dingnuts on Monday September 29, 2025 @04:07PM (#45418148)
> if you don't know how to isolate them in a safe container

Containers are not safe, they are not real isolation at all, Simon! This is awful misinformation coming from someone with your reach.

Containers are just chroots which are a single variable that can be changed to escape the container and if you are running Docker then you have Claude running as root on your system as soon as it escapes the container. They are not real isolation at ALL.

PLEASE stay in your LLM influencer / Python Web dev lane or be much more careful about the technology that you're recommending.

Containers are not for security isolation and LLM output is untrusted input that must be executed in a PROPER context and only after input validation, just like user input.

Re:Claude Code 2.0 by simonw on Monday September 29, 2025 @04:16PM (#45418233)
So tell us how to safely run this stuff then.

I was under the impression that Docker container escapes are actually very rare. How high do you rate the chance of a prompt injection attack against Claude running in a docker container on macOS managing to break out of that container?

Re:Claude Code 2.0 by simonw on Monday September 29, 2025 @04:40PM (#45418504)
I'll also point out that I've been writing about security topics for 22 years. https://simonwillison.net/tags/security/
Re:Claude Code 2.0 by th0ma5 on Monday September 29, 2025 @04:50PM (#45418607)
> So tell us how to safely run this stuff then.

That's the secret, cap... you can't. And it's due to in band signalling, something I've mentioned on numerous occasions. People should entertain the idea that we're going to have to reeducated people about what is and isn't possible because the AI world has been playing make believe so much they can't see the fundamental problems to which there is no solution.

https://en.m.wikipedia.org/wiki/In-band_signaling

Re:Claude Code 2.0 by tptacek on Monday September 29, 2025 @07:17PM (#45420053)
Seems pretty glib. Be more specific about what "can't" be done? The preceding argument was about the inadequacy of namespaced shared-kernel containers for workload isolation. But there are lots of ways to isolate workloads.
Re:Claude Code 2.0 by dang on Monday September 29, 2025 @04:58PM (#45418677)
Please don't cross into personal attack no matter how wrong someone is or you feel they are.

https://news.ycombinator.com/newsguidelines.html

Edit: We've had to ask you this more than once before, and you've continued to do it repeatedly (e.g. https://news.ycombinator.com/item?id=45389115, https://news.ycombinator.com/item?id=45282435). If you don't fix this, we're going to end up banning you, so it would be good if you'd please review the site guidelines and stick to them from now on.

Re:Claude Code 2.0 by tuesdaynight on Monday September 29, 2025 @05:32PM (#45419009)
Amazing work, dang. Is there a way to report a comment to the mods? Or the flag feature does that already?
Re:Claude Code 2.0 by dang on Monday September 29, 2025 @07:39PM (#45420216)
Flagging, plus in egregious cases you can email us at [email protected].