Tor is a better comparison, especially with the reason why most people consider it highly risky to run an exit node.
In both cases there mere existence of the technology helps both good and bad parties to achieve their aims, that is all that matters.
Now, suppose that there’s a transaction chain which involves you – say said local dealer buying from one of the few other people using it — and how you’d prove that you weren’t a knowing participant in some sort of laundering scheme or deal. This is why tumblers aren’t likely to be used by most people and why running a Tor exit node is so risky: you have to worry about being charged with the worst thing done over the network and proving that you aren’t just trying to look like a naive user.
I think that the ZK (ZCash, AZTEC, etc) technology will solve this problem and we will be back to the case where the government wants to make the technology itself go away.
In reality there is an ongoing battle over encryption where various governments try to ban it, people fight back, etc. I suspect that we will see the same thing play out over cryptocurrency.
In any case I don't see anyone on hacker news asking how we can design an encryption algorithm that prevents all and only the bad uses of encryption. It is generally understood to be a dumb question because good and bad are complicated enough that we cannot hope to formalize them right now, and ad-hoc attempts to make a technology only work for the good guys have been repeatedly shown to help the bad guys (e.g. backdoors etc)