Correct, KYC doesn't stop money laundering or terrorist financing. However, KYC does make it harder by the logic of requiring the users to identify themselves to transmit funds.
I understand this point in the world of fiat, because in this case a real risk exists that criminals end up with seemingly "clean" money in the case of "cooperating" banks/fintechs, and I think in this case KYC/AML requirements are mostly reasonable. In a crypto setting (crypto-to-crypto, like mixers and exchanges) the KYC requirement however doesn't help because of the point I'm making in the OP - mixers don't deliver "clean" money, only "crypto-assets of unidentifiable origin", and atomic swaps and CoinJoins can deliver the same thing.
Identity theft is a whole other topic; that has to do with the multitude of mismanagement of individual organisations rather than the government. GDPR and other similar laws are enacted precisely to tackle this problem.
My opinion is a bit different: the more KYC you require, the more identity theft you'll provoke, because you're opening a new attack vector. And there will always be a spectrum of "safer" and "unsafer" services. It's possible that many organizations are handling data security well. But there will be always an arms race between blackhats and these organisations and even those handling it well today may lose tomorrow. Software is never perfect, the problems of bugs and vulnerabilities haven't been "fixed one for all time" even 50 years after personal computers became omnipresent, and cybercrime is at least as successful as it was 10-20 years ago despite of technological progress and increased quality management. I think it's as easy at it sounds: The more organizations are required to store users data, the more hacks there will be, even if the "error rate" was falling due to increased requirements (e.g. due to GDPR-style laws).
You can expect proportionality to be thrown out in court for cases involving surveillance or national security, as historically seen by v. NSA for surveillance related lawsuits.
That's again why it's important to act against these excesses.
A ban on conventional mixers doesn't result in better opportunities for criminals to launder their funds; atomic swaps or other anonymizing methods were always available to all and it is widely and easily accessible for everyone.
I think the logic expecting an increasing anonymity set/degree for decentralized privacy tools is reasonable if the number of (trustworthy) mixers decreases. Liquidity above all for atomic swaps is still low.
For the record, I'm not advocating against privacy. Giving basic rights to the citizens seems to be detrimental in the eyes of the government and if they are able to make money laundering harder, then I don't think they would be reversing their decisions anytime soon.
Yep, I didn't interpret that you were actually against privacy
However, the consequence of your argumentation is that you're ok with a "two-tier" crypto ecosystem: a KYC-bound "legal" tier and a "grey" tier composed of "unregulated" mixers and offshore services. The problem is that this "grey" tier is increasingly driven into a darker shade of grey due to increasing legal attacks, and also is getting bigger (see Samourai's Whirlpool, which should normally a perfectly legal "messaging service" as afaik it never signs transactions, but is now under attack), without that helping with the fight against crime, but mainly impacting normal users looking for privacy.
This is why I wrote before that the status quo is unsatisfying. I could have agreed with maintaining the status quo five years ago (before e.g. AMLD6 was a thing) but I think we'll not return to that if a mentality change isn't happening.
(This also answers the problem I have with your last paragraph).
KYC and AML policies has resulted in banks being able to detect and stop these transactions. These are policies proven effective in the real world, and whenever it seems to fail, they punish the banks and smack them with a lawsuit. I bet this would happen if you were to legalize mixers. My question is: How do you answer the points about how criminal proceeds has been funnelled over through mixers, and the steps taken towards stemming them. Given that these aren't detectable, save for the large ones, how should we ensure that the principles of Bitcoin is upheld (no censorship, fungibility, etc).
We could imagine a policy of "either KYC or delay and react". Basically similar to the recent regulations of harassment in social media.
Mixers could be required to provide a communication channel for freezing requests. In addition to authorities (=police) in an ongoing investigation, if a Bitcoin/crypto user can demostrate that coins were stolen or taken as a ransom (signing a message with the private key used for the transaction), then a "cooperating" mixer has to retain this information and freeze funds which are directly connected with them. In addition, this class of "minimally cooperating" mixers could be required to delay all funds for an hour (close to the 6 confirmations) before they are transferred to the withdrawal address. We could even think about a protocol which includes a timelock and an IF_ELSE clause in the withdrawal transaction, where the user would only be able to move the funds after several hours, and previously the mixer could transfer it back. Taproot would allow that in a reasonably private way.
Such a policy would lead to time pressure for the criminal. Of course the criminal would then probably try to CoinJoin/atomic swap before, but the mixer's importance in the "laundering" process would then decrease. Most mixers of this class would never be used by criminals or only by the "small fishes", and thus could concentrate on the legal "privacy" market. For someone wanting to increase the privacy of his coins, the "delay" process would not be a problem.
The mixer would be liable only if he fails to demostrate that he did everything to try to freeze these funds: react in time to the "freezing request" and/or not implementing the "delay" protocol.
This is of course only an idea, and the work to convince the authorities that his may be a better approach would probably be very hard, but perhaps it can be done if public pressure (from voters) goes in that direction.