For now, I'm assuming they have good protection against automated solutions. If it were just a matter of detecting plagiarism, you're right that it would be trivial. (Heck, I even cofounded a company that did code plagiarism detection, so I have a very good sense for how easy or hard it is. tl;dr: it's straightforward to defeat a plagiarism detector if you are sufficiently motivated and knowledgeable and willing to put in some time to go through all the code, but people/companies who plagiarize are typically none of those.)
Detecting whether someone else is doing the candidate's work, on the other hand, is impossible. Unless you bend the rules. Fortunately, the rules seem pretty pliable to me -- when a candidate arrives at a company's interview, spot-check them on their understanding of a few of "their" trickier solutions. It's a logistical nuisance, to be sure.
The verification burden shouldn't be on the hiring company, it should be Kalzumeus...
I don't think they'll really need to worry about these things for quite a while, though. Good problems to have, I guess.
It also depends on the type of problem. If you just need one weird trick to pass the level (e.x. use this SQL injection) then that's easy to game. If you have to write a bunch of distributed systems code (e.x. Stripe's last CTF) that's a lot harder to cheat.