Aside from running a native application in a VM, most other sandboxing solutions are escapable, mostly because no mainstream desktop OS was architected with san

Aside from running a native application in a VM, most other sandboxing solutions are escapable, mostly because no mainstream desktop OS was architected with sandboxing in mind. Jails are as good as we got, and even then the kernel is far too insecure IMO. Its very hard to support a POSIX api and then retrofit sandboxing. And with VMs, there are numerous escapes, its just another frontier of security issues, not unlike browsers. But browsers were built with sandboxing as a core part of the design, and if you were to not use the newer, far buggier APIs, like Bluetooth and WebGPU, I'd argue that the browser would be relatively safe for a layperson. And much more practical than a VM.