I think the point is that applications can sandbox (parts of) themselves in a way that works across distros and doesn’t require a specific system configuration. Think of BSD pledge(), or browser sandboxes. (Browser sandboxes currently use seccomp on Linux. seccomp is good for when you want to disallow almost all syscalls, which browser sandboxes can do due to their highly elaborate design, including routing all file and network access through another process. It’s not so good for when you would like to keep interacting with the kernel directly but just want to limit accessible file paths.)