[TZubiri]

I never get tired of posting this answer because everyone on the internet is adopting this hot take:

If you look at it with your eyes crossed, Anthropic and the chinese are doing the same thing.

If you look at it with nuance 1 the chinese are doing way worse stuff, and 2 stealing from a thief would still be stealing

1. The chinese are making multiple accounts (at least 49,000)[1][2], using proxies/VPNs, possibly using residential computers and infected computers (unless you think the chinese are doing due diligence to ensure their purchased IPs are kosher). All accounts need to be created with a real name, and especially so if the paid models need to be accessed and paid with a credit card. So this is beyond IP theft and getting closer to fraud. These are all techniques that are well studied because they are used by criminals and cybercriminals, textbook stuff. Consider if that was not sufficient, that China is banned from using the product, so they need to use identities and locations not just to avoid relating the accounts between themselves, but merely to allow account creation. What identities are they using to create accounts.

Compare this to Anthropic which reads notes made a deal in an IP theft case paying billions because they bought books and scanned them but buying the books wasn't sufficient retribution for the authors. Or that they gasp scanned the internet, like Google.

Not having nuance to see the difference between the two companies is something I expect of the twitter echo chamber copying hot takes for upvotes, not hacker news.

[1] https://arstechnica.com/tech-policy/2026/06/anthropic-claims... [2] https://www.anthropic.com/news/detecting-and-preventing-dist...

[bildung]

What seems to be missing from that take is that a) Alibaba paid for the access b) there is no IP theft because LLM output is not copyrightable.

Anthropic seems to want to both own and eat its stolen cake.

[Obscurity4340]

Would be an interesting exercise to have the frontier models calculate their civil liabillity or the extent of the liabillity the can impose on fellow pay-it-forward theives

[codedokode]

First, LLM is merely a tool and its output belong to whoever generated them. If a Chinese researcher used their creativity to generate a response, the copyright belongs to them and AI companies have no rights to it. Second, Chinese release many of their models for free, thus being on a noble mission to make AI available for every country (unlike certain company whose promises were nothing but words). For comparison, US companies do not release anything and want to keep AI for themselves and decide who gets to use it.

> stealing from a thief would still be stealing

Stealing from a thief hurts thief industry which is a win for society.

> The chinese are making multiple accounts

Not a crime. AI companies also ignore robots.txt and applicable laws when illegally copying copyrighted material from websites to their servers without author permission.

[TZubiri]

>Stealing from a thief hurts thief industry which is a win for society.

You are welcome to study the law of any country. A crime against a criminal is still a crime.

>applicable laws when illegally copying copyrighted material from websites to their servers without author permission.

If the material is distributed in http without authentication, isn't that sufficient authorization from the distributor? I would think the search + web crawler era would have set plenty of precedent for this.

>Not a crime. AI companies also ignore robots.txt

Breach of contract is not a crime, agreed.

How about identity fraud (accounts by identity proxy, document KYC), computer crime (C&C residential proxies), conspiracy.

And after the June US directive to suspend Chinese access, smuggling, false statements to regulated entity.

These are all criminal charges that are presumably not levied because of the adversarial relationship between those countries. But if this happened in the US you would probably be seeing at least a civil claim and potentially criminal charges. Hell if this were in any other western country you would see the same. Consider CloudFlare vs Spain, much lighter criminal accusations, and there's already a criminal investigation brought where the CF CEO is indicted.

Non-trivial lack of nuance when you can distinguish between a domestic civil case and a criminal international case between 2 world powers with great judicial tension.

[xpct]

Let's not sane-wash Anthropic's book theft. No, they didn't just 'scan' the internet, they created a tool for worldwide license washing and got fined an insignificant amount for it.

[TZubiri]

You may be conflating the book thing with internet scanning.

On the book case, a class action case was brought to court and it was settled. There's no use in bringing it up further, it has been settled, and it bears no relation to the Anthropic v China case.

You like programming? Think of encapsulation, imagine if you had to think about f(x) but someone brings up y, now you have to think about f(x,y) and what other parameters might bear relationship? The law simplifies by compartimentalizing. And it doesn't even bear a tradeoff, judgment(case1,case2) isn't better than judgment(case1)+judgment(case2).

[xpct]

My response was directed to your insincere characterization of Anthropic's actions. As we can see from the comments here, the public opinion hasn't settled the same way as the court case has, and that's why it's still discussed.

[TZubiri]

But 'the public opinion' doesn't matter. The case was brought by copyright holders to the courts, and Anthropic and the copyright holders made a deal wherein the copyright holders were paid money in exchange for dropping the case and all claims to the intellectual property of Anthropic's derivative product.

If the damnified have considered the matter settled, why would it matter what third parties have to say about it? Third parties would have pushed for more compensation, or ownership of the derivate product. If you feel damnified yourself you can open a case and explain why the actions of Anthropic have hurt you personally.

Otherwise it is a matter that doesn't concern the general public, we have no say in it and there is no right to be offended on behalf of parties that have already settled the argument.