Anthropic Bans Claude Access for Any Company Majority-Owned by Chinese Capital
Event Review
On September 5, 2025, Claude's developer Anthropic suddenly issued a major announcement — "Updating Restrictions of Sales to Unsupported Regions."
The core content: Prohibiting companies whose majority equity is held by Chinese capital, and their subsidiaries, from using Claude services, regardless of which country the company is registered in.
This is not a "grayscale test" but a hard policy effective immediately. On the day of the announcement, a large number of Chinese enterprise users found that APIs could not be called and accounts were disabled, with no transition period or grace period.
How Strict Is the Policy?
Anthropic's regulations are very clear: as long as a company is directly or indirectly held by a Chinese entity with more than 50% equity, it cannot use Claude services.
Specifically including:
- All companies registered in Mainland China
- Subsidiaries registered overseas but with a Chinese parent company (such as ByteDance International, Tencent Overseas, Alibaba Cloud International, etc.)
- Chinese enterprises using Claude through cloud service relays
- Startups with Chinese investment backgrounds (if Chinese investors hold > 50% equity)
In other words, even if your company is registered in the United States or Singapore, if Chinese capital dominates the equity structure, usage is prohibited.
It's not just enterprises affected. ByteDance's AI code editor Trae overseas version (registered in Singapore) is also caught in an awkward situation because of this.
Reasons Given by Anthropic
Anthropic mentioned three core concerns in the announcement:
1. Legal and Compliance Risks
Anthropic believes that the laws of authoritarian countries may force companies to share user data and technical details, and Chinese enterprises may be required to cooperate with intelligence agencies.
2. National Security Considerations
Anthropic claims this move is to "ensure that transformative AI capabilities advance democratic interests," preventing Claude's capabilities from being used for military purposes or developing competitive AI products.
3. Compliance with US Export Controls
Anthropic uniformly defines countries like China, Russia, Iran, and North Korea as "adversary nations," stating this is a requirement to comply with US export control regulations.
Deeper Underlying Reasons
On the surface, it's a compliance issue, but behind it reflects the trend of AI technology becoming geopoliticized.
The DeepSeek Stimulus: DeepSeek's success has made the US AI industry alert — Chinese teams can build powerful models with fewer resources, making American companies start to worry about their technology being "reverse-engineered" or used to cultivate competitors.
Pressure from Washington: Against the backdrop of the US-China tech rivalry, American AI companies face increasing political pressure. Not imposing restrictions could lead to Congressional hearings; imposing them loses the market — Anthropic chose the former.
Commercial Gaming: Anthropic's competitor OpenAI had previously implemented similar restrictions on China. Anthropic following suit is, to some extent, a manifestation of industry "consensus."
Impact on Chinese Developers
Enterprise Users: The impact is enormous. Businesses relying on the Claude API need to migrate urgently. Some enterprises reported service interruptions without any warning, directly affecting online business.
Individual Users: The policy targets "companies and entities," so personal accounts are theoretically unaffected. However, due to the GFW, individual users in China have always needed special network means to access claude.ai, a barrier that already existed.
Domestic Alternatives
The good news is that domestic manufacturers reacted quickly:
- Zhipu GLM-4.5: On the day of Anthropic's announcement, it launched a "Claude Migration Plan," fully compatible with the Claude API protocol, requiring only 3 lines of code modification to migrate, at a price only 1/7th of Claude's.
- Alibaba Tongyi Qianwen: Enterprise-level API service, supporting 200+ languages, with complete compliance support.
- Baidu Wenxin Large Model: Has advantages in Chinese understanding and generation.
- Tencent Hunyuan Large Model: Deeply integrated with the Tencent Cloud ecosystem.
Final Thoughts
Anthropic's restriction policy sounds an alarm for all Chinese developers: Core technology cannot rely on others.
Whether an enterprise or an individual, one needs to think about a question — if an overseas service you depend on suddenly becomes unavailable one day, do you have a Plan B?
In the long run, this event may accelerate the autonomy process of the domestic AI ecosystem. When external dependencies become unreliable, building one's own capabilities becomes the only choice.
What do you think of Anthropic's policy? Welcome to discuss in the comments section.
Top 12 of 27 from juejin.cn, machine-translated. The original thread is authoritative.
Should look more at our own reasons, why we're grouped with Russia and North Korea.
Tell me why then, I've got my little stool ready.
No matter how you think about it, it's just that way, because you don't fit in with others.
Not thorough enough, just ban questions and answers in Chinese directly and that would be a killing blow [lightbulb moment].
That would be pretty much like racial discrimination, it would face boycotts!
In that case it would become a 'backward' model, and there'd be no need to use it anymore [eating melon spectator].
So the Earth needs to be unified [simple smile].
After unification, unifying the language is very necessary.
Unify the currency too, don't bother with all these exchange rates, conversion is a hassle.
He provides services to the US Department of Defense, a boss who worked at a Chinese company, if not anti-China then pro-China.
No wonder, but it doesn't stop Claude from being good to use.
Anthropic is purely a bitch of a company. Looking forward to Saint Liang stepping up.
I've noticed a lot of AI self-media or tech people like to chat about this topic, actually tech people aren't suited to discuss it, because this isn't a tech issue at all. All that compliance, rules, and clause jargon is just to have a righteous pretext and seize the moral high ground. So Sudan, Palestine, Haiti, Syria and other refugee nations are fine? Well, that figures, these countries are struggling to survive, so they definitely have no democracy issues, pose no threat, and can't possibly be used to develop competitive AI products. The day we fall apart and become a refugee nation losing our threat value, we can also be removed from the list. Or learn from Japan and South Korea, invite a military presence + sign a Sino-US Plaza Accord and this problem can also be easily solved. After all, hostile nations are unilaterally determined by the US, they say you are, and you are.
Isn't all responsibility on Baidu?
You really have a big problem with Baidu, huh [grin].
2
This was bound to happen sooner or later, the pressure from business and compliance is right there. Fortunately, there are already quite a few alternatives domestically, DeepSeek, GLM and these are already very capable on programming tasks, plus API relay channels are also improving, the impact isn't as big as imagined, just adapt the workflow a bit.
Not bad.
DS, go for it, the model's capability is my capability, can we catch up to GPT-5.5, Claude 5?
This is actually a good thing.