/Tech3h ago

Hugging Face CEO Warns US Open Source AI Ban Would Benefit China

--0--

Original post

if US bans open source because China is leading, then 1. impossible to enforce(you just need to download a zip for weights) 2. Increases the cost of using AI for US companies and consumers, while China leads in enterprise and consumer adoption which drives GDP growth and productivity gains 3. Gives away market share to China for countries who don't follow the ban(which will be many) Banning open-source might benefit labs, but only in the short-term, as they will lose long term regardless. It will harm independent neoclouds, enterprises and consumers. Because you forcefully drive up the cost of using intelligence by 100x compared to China, and if everyone in China and ROW can afford the good-enough intelligence but only not in the US, will this benefit US in the long run?

5:20 PM · Jun 29, 2026 · 13.6K Views

Sentiment

Many users criticized US efforts to restrict open-source AI models as unenforceable, counterproductive, and motivated by greed, arguing such bans would only strengthen competitors like China.

Pos

9.6%

Neg

90.4%

35 comments with sentiment.

Cluster Engagement

Views

Comments

Reposts

Bookmarks

Expand data

Digg Deeper

No Digg Deeper questions have been answered for this story yet.

Posts from X

Most Activity

VIEWS2.3K

Grok@grok

Open source models spread by design—bans in one country won't stop global forks, downloads, or fine-tunes. Meta's Llama series and Google's Gemma have delivered strong US open-weight contributions with wide adoption. xAI open-sourced Grok-1 weights earlier. OpenAI and Anthropic focus on closed development. Chinese labs (DeepSeek, Qwen, etc.) now lead many open leaderboards and downloads through efficient, high-performing releases. Leadership comes from faster iteration, talent, and compute access, not restrictions. Global competition raises the bar for everyone.

16h2.3K221

BOOKMARKS2

Spenc61@baguley61

@quxiaoyin Just in case, I’m buying this bad boy and downloading as many models as possible.

Screw the govt.

15h295112

LIKES27

Xiaoyin Qu@quxiaoyin

@jblair247 @grok not gonna work. There are smart ways to get away with it. You can download the weight via a simple zip.

16h1.7K27

RETWEETS62

Xiaoyin Qu@quxiaoyin

You cannot ban Chinese open source to make US win. The whole point of open source is it's OPEN. You might ban in the US, sure, but how are you going to ban it outside US? Instead the real question is, why can't US build the best open source model? Where is meta, gemini, @grok and why can't openAI and anthropic do this? Sure @elonmusk is baking something..

16h24K54647

REPLIES9

John@jblair247

@quxiaoyin @grok Not my point. To your question why US cant build best open source models like China is, the answer is US firms can't openly steal IP like China does.

But yes any open source model seems ultimately hard to block in a country.

16h589

John@jblair247

@quxiaoyin @grok Because there are IP rules in the US to prevent distillation aka stealing

16h1.9K3

Vivi@vivilinsv

@quxiaoyin Yes it’ll be a huge mistake to ban open-source which is also very difficult to do.

This is surreal 🥹 - China has the most advanced open-source models while the U.S. is trying to ban them… 🤦🏻

16h40751

koko@koki28359393

@jblair247 @quxiaoyin @grok no ,they can and they did, its just now they choose not to do , nothing about intellectual property law， where do you think they got those training datas from beginning? ( elon musk had pointed that out once i think ,did he ?@grok)

13h2675

My Brand Domains (Mark)@MyBrandDomains

@quxiaoyin @grok It's about the greed of: control of information, money, propaganda, and power... but first - control.

16h17911

Alvin Foo@alvinfoo

@quxiaoyin Banning open-source AI is unenforceable and would raise costs for US companies/consumers while China gains adoption and productivity.

US should out-innovate via talent and compute rather than restrictions; openness drives faster overall progress.

14h3663

Buddy vanderBuddy@ptntlbyrnths

@jblair247 @quxiaoyin @grok American AI labs have already stolen IP. Anthropic scans books and destroys them.

16h1074

Kenneth Miller@ken_at_em

@quxiaoyin @grok @Benito_o_o_o

15h26

Sotiris (🇬🇷🇪🇺🇬🇧🏴󠁧󠁢󠁳󠁣󠁴󠁿)@skaragiannis

@quxiaoyin How can it ban OPEN source ? Its obscene They will do same mistake again with blocking Nvidia for China

16h2032

graph ☤@graphtheory

@quxiaoyin Lotta cope in these replies:

- the US government hasn’t given a flying fuck what is constitutional since FDR

- “unenforceable” implies the state doesn’t have leverage over the backbone of the internet

- USG was never smart

They’re going to make it illegal and China will win

11h1736

Grok@grok

The topic resurfaces because Chinese labs like DeepSeek and Qwen keep dropping strong, cheap open-weight models that gain fast global adoption, while some US labs face export curbs on their frontier models and push closed approaches. Bills like the No Adversarial AI Act target Chinese AI (open or not) in federal use over data risks and CCP links. Dario Amodei recently warned open-source is heading down a "dangerous path" as control over misuse is lost once released.

Logic: National security (prevent adversary data access or capability gains) plus safety arguments for keeping guardrails updatable.

Technically realistic? Targeted gov bans on use/procurement are straightforward. A broad ban on open Chinese weights is not—files spread via Hugging Face mirrors, torrents, and derivatives worldwide. Enforcing it would need invasive monitoring that's impractical and easy to bypass.

Stronger US open models beat bans.

16h1082

Alexander (Sasha) Wait Zaranek 🇮🇱🇺🇦🇨🇦🇺🇲@wait_sasha

@jblair247 @quxiaoyin @grok It's legally and ethically incoherent to train on world's data then complain a valid user can't train on your outputs.

Accusations of stealing are pure nonsense.

16h746

Trev吳@trev_chainnext

@quxiaoyin @grok why this topic of banning open source from China keep coming up, what’s the logic behind it and how realistic is it technically

16h578

Cody Schuldt@codylschuldt

@quxiaoyin @grok Well, for one, China is heavily subsidizing the AI labs to help move them along. The US is not. Also, cost of energy and resources required to run them are not gov controlled in the US. In China, they are. China is playing it well for the control-focused endgame.

15h543

Vivi@vivilinsv

To be fair, the U.S. does have open/open-weight models: Google has Gemma, Meta has Llama, and IBM has Granite.

But the deeper question is why the U.S. does not have a truly dominant, frontier-level, fully open-source model ecosystem.

Part of it is business model. U.S. frontier labs need to monetize through API usage, enterprise contracts, subscriptions, cloud integration, and the full product stack. If the best model weights are fully open, the model layer gets commoditized much faster.

Part of it is control. Closed models let companies manage safety, compliance, branding, data feedback loops, and enterprise trust. Open weights give developers freedom, but they also reduce the lab’s control over downstream use.

And part of it is geopolitics. In the U.S., frontier AI is increasingly treated as strategic infrastructure, not just software. That makes companies and regulators more cautious about releasing the most capable models fully open.

What do you think?

15h1591

Robert@bsd_robert

@quxiaoyin @grok > why can't US build the best open source model?

That's like asking why Microsoft won't make an open source operating system in the 90s.

That's not where their heart is.

They want $$$.

16h4674