9 comments

[ 3.1 ms ] story [ 28.4 ms ] thread
I have a feeling that China will be OK. Wouldn't these restriction protect the native LLM ecosystem and will give them room and revenues to grow? Strong internal-only competition will still make them competitive.
China can train on pirated books all day long, unlike Anthropic or other Western competitors
My understanding is Anthropic is closed source and has been unavailable for Chinese companies for a long time, and if they wanted to use offshore entities to use closed-source LLMs, they would have gone to OpenAI...?
And people complain the TSA is "security theater"...

Yeah, I'm sure whatever restrictions they put in place will make it so hard for Chinese entities to access their APIs.

This is simply an attempt by Anthropic to pretend to play nice with the current administration (and to be clear, I don't blame them - let's just not pretend that this actually prevents China getting access to Anthropic).

So what is stopping anyone in principle just reselling API access through a middleman?

Should I set up the company now and rake in the billions?

Alas there was no way to pause my $200 Max subscription while I'm in China for two weeks...
The 20 countries that Anthropic won't do business with:

  Afghanistan
  Belarus
  Central African Republic
  China
  Congo, Democratic Republic of 
  Cuba
  Eritrea
  Ethiopia
  Iran
  Libya
  Mali
  Myanmar
  Nicaragua
  North Korea
  Russia
  Somalia
  Sudan
  Syria
  Venezuela
  Yemen
I took their list here[1] and diff'ed it with Wikipedia's list of sovereign states.

They don't accept the Democratic Republic of Congo, but the Republic of Congo is OK. There are indeed two Congos with ridiculously similar names, and apparently the one that calls itself democratic is the bad one.

Anthropic is fine with Taiwan.

[1] https://www.anthropic.com/supported-countries (They give two lists, one for API and another for Claude.ai, but the lists are identical.)

> including adversarial nations like China—continue accessing our services in various ways [...] our commitment to ensuring that transformative AI capabilities advance democratic interests

Ah, right, China, the only country which actually releases open weight frontier models. Anthropic definitely doesn't want Chinese companies to use their APIs because of their commitment to "advance democratic interests", and not because they want to prevent them from distilling their models and releasing them for free for everyone to use. Definitely. Fortunately since this use case is already breaking the TOS I doubt any extra legal maneuvering is going to stop anyone.