/Tech4h ago

AI2's Nathan Lambert warns a major breakthrough in Chinese open-weight LLMs would trigger a comprehensive U.S. ban

Story Overview

Nathan Lambert, known for his work on open models at AI2, posted that a standout performance jump from Chinese open-weight LLMs would likely prompt the U.S. to block the entire category. The reasoning centers on national security priorities overriding any preference for open-source access, even when the models themselves carry no direct ties to restricted technology.

3621173024.9K

#80

Original post

Nathan Lambert@natolambert#80inTech

The only reasonable expectation if you're a fan of open weight models is that if there's a major step in chinese open-weight performance, there's a good chance the whole chinese llm sphere is banned.

National security apparatus will happily give a big "fuck you" to open models.

Nathan Lambert@natolambert

Threading the needle in this post of anthropic has done some bad things for AI governance & the discourse but the actions of this administration are way worse so we need to get a handle on it before stronger models, open or closed, come along soon. https://www.interconnects.ai/p/welcome-to-the-agi-era-of-ai-governance

11:07 AM · Jun 14, 2026 · 18.9K Views

Policy Risk

Security concerns reshape access rules

Open-weight enthusiasts would face the steepest limits if the warning holds, since the same rules could sweep up domestic research and startup use cases that rely on those weights today.

Open Question

No policy trigger confirmed yet

A related prediction market puts the odds of any new federal limits on open-weight LLMs by the end of 2026 at roughly thirty percent, leaving the actual timing and scope of any ban as an unresolved variable.

Sentiment

Many users opposed the national security push to ban Chinese open-weight LLMs, arguing it would reduce US leverage, harm American developers, and amount to ineffective protectionism.

Pos

0.0%

Neg

100.0%

8 comments with sentiment.

Cluster Engagement

Posts from X

Most Activity

VIEWS2.6KBOOKMARKS3LIKES28

Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex

Obviously this isn't a game, this isn't even a business the open weights consoomer isn't important for the State

Nathan Lambert@natolambert

The only reasonable expectation if you're a fan of open weight models is that if there's a major step in chinese open-weight performance, there's a good chance the whole chinese llm sphere is banned.

National security apparatus will happily give a big "fuck you" to open models.

3h2.6K283

RETWEETS1

Nathan Lambert@natolambert

Goes without saying I think this is a bad idea.

Nathan Lambert@natolambert

The only reasonable expectation if you're a fan of open weight models is that if there's a major step in chinese open-weight performance, there's a good chance the whole chinese llm sphere is banned.

National security apparatus will happily give a big "fuck you" to open models.

1h1.6K61

REPLIES2

Steve Newman@snewmanpv

@natolambert @nv_pavlichenko I thought Huawei GPU sales were generally thought to be limited by manufacturing capacity, not demand? (Noting that the state is taking actions to promote demand.)

2h251

Nathan Lambert@natolambert

@CFGeek im not in the gambling business around my decisions. Having watched the govt closely, they're massively amplifying the anti chinese model rhetoric and I expect it to continue this year.

Charles Foster@CFGeek

@natolambert That would be a huge move. I think it’s pretty unlikely that the US will restrict import of Chinese LLMs for ordinary use in the near future.

You might be interested in this play-money market: https://manifold.markets/CharlesFoster/will-the-us-government-introduce-an?r=Q2hhcmxlc0Zvc3Rlcg

3h16741

Nathan Lambert@natolambert

@CFGeek US is most likely to restrict import bc its govt is the furthest along in dealing with AI and forming strong opinions about it

Charles Foster@CFGeek

@natolambert Are you saying that other governments (US, Europe, etc.) will restrict “import”/use of Chinese LLMs? Or that China will restrict its LLMs domestically? Or maybe that China will restrict “export” of its LLMs?

3h40950

Nikita Pavlichenko@nv_pavlichenko

@natolambert Why not CCP banning releasing open-weight models first if they take the lead or close to it?

4h134

Charles Foster@CFGeek

@natolambert As a policy person, I’m similarly watching what the USG is doing. I haven’t seen any indications that they’re about to restrict access or use of Chinese models. Maybe you’re seeing or hearing stuff that I’m not!

2h431

Nathan Lambert@natolambert

@nv_pavlichenko I don't think the Chinese LLMs will ever lead in the next 1-3 years. We will see signs of it potentially coming by absolutely insane numbers of Huawei chip sales

4h1166

Charles Foster@CFGeek

@natolambert That would be a huge move. I think it’s pretty unlikely that the US will restrict import of Chinese LLMs for ordinary use in the near future.

You might be interested in this play-money market: https://manifold.markets/CharlesFoster/will-the-us-government-introduce-an?r=Q2hhcmxlc0Zvc3Rlcg

3h72

Pehdrew@Pehdrew_

@natolambert @natolambert we'll need this but for LLM training.. @PrimeIntellect any updates?

3h49

Josh@JoshPurtell

@natolambert IIUC USG wanted Anthropic to either patch the guardrails providing unfettered access, or give them more lead time to harden against unfettered access.

But presumably if Chinese models remain 3-9 months behind US, there will always have been sufficient lead time

3h791

Kavin Stewart@kavinstewart

@natolambert I can see it, but it’s a strange position. kinda like saying “only foreigners and the govt are allowed to have nukes, not citizens”

3h701

TG@ossaijaD

@natolambert Does ai model need to be inherently threatening to power? How else can ai be engineered so it can be whole embraced like other technologies?

3h70

selv Marū Atma-Goma@MagisterLudiX

@natolambert Not here in Europe, I guess.

4h64

Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex

@nv_pavlichenko @natolambert There's very little gain in banning open source if you're merely close, and they already have been close. From the CCP's perspective, R1 was frontier. They congratulated DeepSeek and asked for more of the same

3h52

Kyle@kyle_mccleary

@natolambert Maybe, but actually enforcing this ban would be almost impossible, don't you think? Maybe for enterprise use it's doable.

3h41

Plastic Soldier@PlastiqSoldier

@natolambert I think I'd expect China to move to closed weights instead.

3h39

Didymus@BasilEsq_

@natolambert I don’t see why they need a major step change before USG bans them. We’ve banned BYD and Huawei and neither of those can make bio weapons.

3h22

Android@AstralBIO

@natolambert They literally would have to come pry qwen of my MacBook. Come and Take it!

2h21

roanoke_gal@roanoke_gal

@natolambert What's the best steps we can take to prevent this from happening?

3h16