We've also known since 2007 that "you can't hide secrets from the future with math[1]." So, if you're doing anything illegal and that information is available to a network, it's a matter of time, not encryption.
Also worth pointing out, in the video he reminds us that the definition of "crime" is not set in stone.
If the time that it will take to break the encryption is significantly longer than your expected remaining lifespan (and anyone you care about), then you probably have nothing to worry about. In any case, sneakernet is an option that is always available.
These estimates of security should be considered an upper bound, not a lower bound or average. In other words,
> If the time that it will take to break the encryption by brute force is significantly longer than your expected remaining lifespan
....just because we don't know about a vulnerability in SHA-3 today doesn't mean that we won't in ten years. SHA-1 and SHA-2 were once thought to be secure, and they may yet be even more broken than we realize.
Furthermore, that also makes a major assumption about computing hardware. Particularly with the (possible) advent of quantum computing and the possibilities that large-scale quantum computing would provide, I think it's impossible for anybody to do more than speculate about the future like that.
Also worth pointing out, in the video he reminds us that the definition of "crime" is not set in stone.
[1] http://en.wikipedia.org/wiki/Secrets_from_the_Future