As having some experience on military crypto, as Aleksander Rassasse said, everybody relies on Kerckhoffs's principle. Yes, everybody relies on.
However, there are some differences. Militaries still think that even we relay on the Kerckhoffs's principle, as a security measure, we will not release our cryptographic algorithms, Error-Correcting Codes, protocols, key distribution systems, etc, and the crypto devices to 3rd party investigation.
Instead, closed communities try to imitate the open community. That is; setting up groups that have different aims as algorithm designer group and analyzer group. Former group design and the latter requires some time to analyze. The reason is simple; the path of wisdom is unique.
This is completely personal taught; (Also, I believe that NSA or other agencies, try to outsource some of their problems so that they can use the result in their problem)
As a consequence of this kind of restrictions, fewer minds analyze and see the problems. Whereas, in industrial, at least somebody can buy the device and reverse engineer it, as in A5/1, and finally show that it is breakable.
Military actually benefit from these restrictions. As an example, when you listen over a radio channel, one cannot distinguish whether the signal just an open message+ECC or message+zip+ECC+encryption, etc.
One interesting example of fewer mind analyze is the SIMON cipher by NSA. Now, it is almost broken.
One interesting example is in OpenSSL, Dual_EC_DRBG. This is missed 7 years from the open community.
Conclusion; let them play their games, and let play ours.
I was once part of a closed one.