So, I should obviously link to My Weirdly Complex Feelings About US Anti-Money Laundering, which is retracted in full.
So, the comparatively minor errors first: KYC laws can be subject to abuse, and there's pretty strong reason to believe that all manner of legal activity is denied banking services because of them, notable legal drug sales and the adult industry. So far as I can detect, there's no problem with banking these people in principle, but the laws enable the federal government to apply pressure and exclude them. That's wrong, and it should have been more obvious to me that this was connected to AML/CFT.
Also, I mentioned that AML/CFT had a comparatively light footprint, but in truth the footprint is bigger than you might guess – there's reason to believe it's a meaningful cause of the transaction delays modern infrastructure is just now getting past. Obviously some of that is fraud prevention, and some of it is counterparty risk minimization (a couple flavors of that, really). But according to a new report, there's a lot of speed gained when you just making AML reporting faster, and that was a major benefit of using the Ripple network. So, my bad.
This is the report you should read. Perhaps over my own blog.
The main problem is that, even with dragnet surveillance, the amount of enforcement is extremely limited, and there's no reason to suspect they're even detecting the overwhelming majority of crimes. It does not stop terrorists from transacting in dollars. It does not stop gangs from using major international banks. It doesn't even seem to meaningfully limit the amount of criminal income that can be used to reinvest in future crime. Read that report. It's important.
If this is the high water mark for government surveillance (and I think it still might be), I think it makes a pretty compelling argument for getting rid of essentially all laws in this area and reversing all other government programs that want information about normal Americans. It turns out they can't even use it to stop crime.