What I'm taking away from this is that anything David ever has made or will make in the future should not be trusted.
While I'll grant that the you're partially justified by the ridiculously bad summary, your takeaway is dead wrong.
First, having just skimmed through the article and the (very interesting!) paper, let me point out why the summary is ridiculously bad. Chaum's protocol does not include a backdoor, and certainly not "just to please governments".
What Chaum did was to describe a really cool anonymous routing and communications protocol, with a number of highly desirable properties. The biggest one is that his protocol is designed to be secure against nation state access, unlike Tor. It should also be quite a bit faster than Tor because communications require no public key cryptographic operations; everything is done with very-fast symmetric crypto, building on top of a precomputed homomorphic encryption. Making this scheme work, though, depends on the existence of a trusted third party (TTP).
In general, relying on a TTP is problematic in contexts where there isn't any obvious person or organization who could be trusted. And for a global communications network that will be used by lots of people and which many governments might like to penetrate, and which in fact is specifically focused on trying to prevent penetration by nation states, there clearly exists NO such single party.
Chaum's solution to the problem of how to trust when no one is trustworthy (a common problem in security design, actually) is to distribute the trust (a common solution, though Chaum's implementation is particularly clever). By arranging things so that the TTP role is spread across many different nations, each of which is fairly trustworthy except in particular areas, and selecting those nations so the areas in which they're untrustworthy are different, and designing the cryptography so that any abuse of the TTP role requires willing participation of 100% of said nations, it may be possible to construct a TTP which is trustworthy in the aggregate, even though no individual member is fully trustworthy.
This is a very clever solution to what I would have said is a completely intractable problem.