AI Cold War Begins? Anthropic Accuses China of Training on Claude | Vantage with Palki Sharma

帕尔基·夏尔马：AI冷战打响？Anthropic指控中国AI使用Claude进行训练

The AI war just escalated. U.S.-based Anthropic claims Chinese AI firms — DeepSeek, MiniMax, and Moonshot AI — created thousands of fake accounts and ran over 16 million prompts on its chatbot Claude. The accusation? Using a technique called “distillation” to train their own models. But distillation isn’t illegal — and American firms use it too. So, is this theft, or just competition in a high-stakes AI arms race? Palki Sharma tells you.

AI领域的对抗态势再度升级。美国人工智能企业Anthropic声称，中国的深度求索（DeepSeek）、MiniMax、Moonshot AI等AI公司注册了数千个账号，向其聊天机器人Claude发送了超过1600万次提问请求。

Anthropic的指控核心是：这些中国企业通过一种名为**“模型蒸馏”**的技术，利用Claude的数据训练自家AI模型。但模型蒸馏本身并不违法——美国科技企业同样在广泛使用这一技术。

那么，这究竟是数据窃取，还是高风险AI军备竞赛中的正常竞争？

以下是印度网友的评论：

sleepyjoe4529
Only the West can steal from others apparently LOL

很明显，也就只有西方能随便偷别人的东西，哈哈

Maurisank
you suck man have mercy bro

你也太菜了，兄弟，放过大家吧

SuperDad1793
Anthropic : Stop stealing my stolen data.

Anthropic：不准偷我已经偷来的数据

EagleKnight35
Even Anthropic has distilled information from the internet to train its Claude AI models. Like other major AI developers, Anthropic uses extensive web-scraped data to train its models. US AI chatbots, particularly from firms like OpenAI and Microsoft, face legal accusations of training on copyrighted content. They are accused of synthesizing data from massive internet sources to generate responses, raising significant intellectual property (IP) and accuracy concerns. Even if OpenAI or Anthropic can present enough evidence to show that DeepSeek extracted data from its models, they likely do not have copyrights over the data. There is ongoing debate concerning whether the outputs of AI models, such as those generated by ChatGPT, can be subject to copyright protection.

就连Anthropic自己也从互联网上提取信息来训练Claude AI模型。和其他大型AI公司一样，它也靠大量爬取网页数据训练模型。美国的AI聊天机器人，尤其是OpenAI和微软的，都被起诉过使用受版权保护的内容训练。他们被指控从海量互联网数据里整合信息生成回答，引发了严重的知识产权和准确性争议。就算OpenAI或Anthropic能拿出证据证明深度求索从它们的模型里提取了数据，这些数据本身也未必受它们的版权保护。目前大家还在争论，像ChatGPT这类AI模型生成的内容到底能不能受到版权保护。

LobgaJulius
@EagleKnight35 Check on what the meaning of distillation, distillation is different from training which is what you are describing

你先搞懂什么是模型蒸馏，蒸馏和你说的普通训练不是一回事

feisalboudghene228
Its different, AI original models train by interacting with other users and the you build a model from scratch. Distillating is using AI to train by interacting with other AI's and build a model similar to the one builded from scratch.

这不一样。原生AI模型是靠和用户互动、从零开始训练的。模型蒸馏是让AI去跟别的AI对话，然后训练出一个相似的模型。

tydf229
Sour loser is not good

输不起就太难看了

darlayjones669
Blaming the other player of cheating whenever you're losing the race isn't the solution.

比赛跑不过就骂别人作弊，这解决不了问题。

jaycdp
well the looting was practically the back bone since 1700, Now it is leaching

从1700年开始，掠夺本来就是他们的立身之本，现在只不过换成吸数据而已

vividclue4923
How was Claude trained as well

那Claude自己又是怎么训练出来的？

java2architect
Hapless folk like us give them all our data without knowing it. They are now yelling China stole our data from them. In US that's a legal issue. In China, they do not give a F.

我们这些普通人稀里糊涂就把数据都给了他们。现在他们反倒喊中国偷了他们的数据。在美国这是法律问题，在中国他们根本不在乎这套。

krishnit2k3
@java2architect Yes 100% true- we give them the data and their API learn from us. Sam Altman makes money out of us.

百分百正确——我们给数据，他们的API从我们身上学习，山姆·奥特曼靠我们赚钱。

Liboch
at least DeepSeek doesn’t charge money

至少深度求索不收钱

johnnyBrwn
Lmao, that's "it's not fair you stole from us after we stole from you!"

哈哈，这不就是：我偷你的可以，你偷我的就不行！

labandonaldhock80
US claims are total bs.

美国的说法全是扯淡。

Chess_and_Universe_Astronomy
They are true, but usa itself steals other's data, so.....a thief is trying to catch a thief

话是真的，但美国自己也偷别人数据，所以纯属贼喊捉贼

FreeSportsEntertainment
US is just jealous

美国就是嫉妒了

zimcoder
You can count on the US to blame China and Firstpost to take cheap shots at the same. Its became so predictable and boring.

美国一定会甩锅中国，《印度第一邮报》也一定会跟着抹黑，套路太老套，看腻了。

francoislechanceux5818
China should stop copying and stealing only to come around and claim to be better in this or that

中国别再抄袭和偷窃了，完了还到处说自己这也行那也行

NeoThabo
So what?! Where did those americans get the data to train their models?

那又怎样？美国人训练模型的数据又是哪来的？

manmohan1730
They did by scrap and caught paid $1.5B, do you think anyone can force chinse company to pay $1.5B?

他们靠爬数据，还被罚了15亿美元。你觉得谁能逼中国公司赔15亿？

bluegg2159
From world. China is banned from world ok. See most of country ok. India, Europe to usa had banned chinese from taking data

数据来自全世界。中国被全世界禁止了。大部分国家都这样，印度、欧洲、美国都不让中国拿数据。

adamsaciid4919
@bluegg2159 while americans takes chinese data so its normal china to take also

美国人照样拿中国数据，那中国拿回去也很正常

jishe8957
I've noticed that Indians have a penchant for a grand narrative—that the whole world is against China. However, India itself is being attacked. China, on the other hand, is beating up the United States.

我发现印度人特别喜欢编大故事，说什么全世界都在针对中国。可印度自己才是被针对的那个，中国现在反而压着美国打。

zimcoder
"US AI companies built safety guardrails".. the funnest thing I have heard in quite a while.

“美国AI公司设置了安全护栏”，这是我最近听过最好笑的话。

josh981234
The US has no moats lol

美国根本没有护城河，哈哈

francoislechanceux5818
Not mentioning India or France in such a video 3 days after the AI Indian summit

印度AI峰会才过去三天，你们怎么提都不提印度和法国

adamsaciid4919
Can't win say they stole, say it's national security risk lol

赢不了就说别人偷，再扣上国家安全风险，哈哈

adamsaciid4919
Btw those american ai companies where they get the data

对了，美国那些AI公司的数据又是哪来的？

morbid747
China AI is free and open source. US AI companies hate them because they will lose money to free AI.

中国AI免费还开源，美国AI公司恨死了，因为免费AI会抢他们生意。

MoC-adsf
Exactly what I thought. Anyone could have downloaded then run it. But it might be seen running from the mother LLM

我也这么想。谁都可以下载运行，但可能会被源头大模型检测到。

P-hl5ln
Saw that blame coming ...They will never says chinese are working hard, has engineers and scientists. Knowledge is distributed evenly accross the globe.

就知道他们要甩锅，他们永远不会承认中国人努力、有工程师、有科学家。知识本来就是全世界共享的。

michaelhmar1885
This is Indian News.

这是印度媒体的新闻。

badbad-cat
I wouldn't blame India for this one. This was spinned by USA's AI clowns who are too desperate and exhausted

这事别怪印度，是美国那些急疯了的AI小丑在带节奏。

pigeon-fd5zq
Anthropic founder is doing drama just to gain value and attention and already done public stunt in India AI summit not shaking hand with Sam

Anthropic创始人就是在演戏，想抬估值博眼球。之前在印度AI峰会上故意不和山姆握手就是炒作。

samsamra6887
similar to..chandu ki chachi ne chandu ke chacha ko

有点像绕来绕去的印度顺口溜

boxtears
An AI company that steals user data is accusing China of stealing

一家偷用户数据的AI公司，居然指责中国偷窃

Avatar_2025
Knowledge is universal and belongs to no one. Those who think knowledge is proprietary are either naive or flat out stupid.

知识是全人类的，不属于任何人。觉得知识可以私有的，要么天真，要么纯傻。

soulpraisemedia
How about using fewer chips yet producing great efficiency, spending $6M where others spend billions, is this copying, distillation?

用更少的芯片做出更高的效率，别人花几十亿，他们只花600万。这也叫抄袭、叫蒸馏？

ricnyc2759
Let me tell you what happened: the Chinese built a time machine and traveled to the future and stole Seedance from the US. Not only that: they also stole thousands of high-speed rail from the US in the middle of the night and move it to China

我跟你说真相吧：中国人造了时光机，穿越到未来，从美国偷了Seedance。不仅如此，他们还半夜从美国偷了几千条高铁，运回中国。

Nomadsabin
Remember Anthropic is most hypocrite company

记住，Anthropic是最双标的公司。

JarvisStars-b1f
I am sure the American AI companies are doing the same thing as the Chinese

我敢肯定，美国AI公司和中国公司干的事一模一样。

Gulfnas1
Here we go, after seedance beat and shake them

来了来了，Seedance把他们打懵之后，就开始搞事了

syedmortuzaasifehsan1018
Better than buying a robo dog from the same China and claiming as of own.

总比从中国买机器狗，然后说是自己研发的强。

bobcharles7716
This is fact American AI companies said it is legal AND they are mining other websites to develop their AI. I guess when the US do it it is legal and when China is doing it is sealing!

事实就是：美国AI公司自己说模型蒸馏合法，还自己爬别人网站训练AI。合着美国干就是合法，中国干就是偷窃！

Roy100PKR
Laughing so much over here I can’t breath. Wait one second, the hypocrisy is richer than the irony. Every State in the US wanted to make their own law on regulation of AI, Big Tech hated the idea of any regulation so asked the Trump admin to fight it in court. Big Tech wants the Wild West, no laws to control what they are doing. So they wanted Wild West, now they’ve got it. So why complain about China if what US Big Tech is saying is true? I highly doubt anything coming from America is true. It has been proven time and again now that they lie over everything. Venezuela, Greenland, Tariffs, Iraq, Epstein etc. over and over they lie. You would be an idiot to trust anything they say at face value.

我笑得快喘不过气了。这双标简直离谱。美国每个州都想出台AI监管法律，科技巨头讨厌监管，就找特朗普政府去法院阻挠。科技巨头想要无法无天的自由市场，现在他们如愿以偿了。如果他们说的是真的，那凭什么抱怨中国？我根本不信美国说的任何话。事实一次又一次证明，他们什么都撒谎。委内瑞拉、格陵兰岛、关税、伊拉克、爱泼斯坦案，一遍又一遍地骗人。谁信他们表面的话，谁就是傻子。

继续阅读：人工智能

此文由 三泰虎 编辑，未经允许不得转载！：首页 > 印度看中国 » AI冷战打响？Anthropic指控中国AI使用Claude进行训练