>BEIJING, Oct 28 (Reuters) - China's proposed artificial intelligence export curbs will push a business model favoured by Beijing and the industry's biggest players: keeping valuable technology at home while selling access to it abroad as rivalry with the US intensifies.>The draft rules, released on October 16, distinguish between foreign customers using Chinese AI services and foreign companies obtaining the models, data or expertise needed to operate or reproduce them independently.>"The structure the leadership prefers, with IP and R&D kept onshore and operations and access sold abroad, is the one the market was already converging on," said Alfredo Montufar-Helu, Beijing-based managing director at consultancy Ankura.>The new rules would broadly allow overseas users to access Chinese AI through websites and cloud platforms while subjecting some transfers of underlying technology or deployments overseas to government approval.> Yang Jingru, a partner at law firm Jincheng Tongda & Neal, described the emerging approach as a counterpart to US restrictions on China's access to advanced technology and investment.>"The US controls the inflow of computing power, hardware and capital, while China is beginning to systematically manage the outflow of models and technological assets," Yang wrote.https://www.reuters.com/business/retail-consumer/chinas-new-export-curbs-draw-line-around-its-ai-crown-jewels-2026-10-28/
Okay
>>110187405OH HELL YEAH!!!
>>110187405So they're just going to stop releasing open source models?Its good for them since that's the model american AI companies have been using to guard their trade secrets.This could go a number of ways.
>CHINA TO BAN OPEN WEIGHT MODELS (OUTSIDE OF CHINA)hell no.>draftcorrect, and the article should've been done and done here.>October 16yep, this is literal old 'news', and was covered by actual business and news sites that very day. but nice try anyway Reuters. >consultancy >partner at law firm Who are these people? Who the fuck cares about the opinion on this from a bunch of randoms? Fuck consultants and analysts anyway. And this is the beauty of Chinese politics by the way. You'll hear about something the exact moment it's enforced, and not a second earlier, which logically means this document is a complete nothingburger.
>>110187405>worse models>can't use them on your own shitlol ok good luck
That's the strength of their models though. Sad times if the only open models are fucking Gemma and Mistral.
>>110187405I'll bet Chinese models will get leaked anyways.The Chinese lax attitude towards copyrights cuts both ways.
>>110189029>worse modelsSeedance 2.5 shits all over anything Americans have made. Even Minimax H3 is better than anything released by Americans.
>chang models start falling way behind>"uh actually they're too good so we're not going to be releasing them anymore"
>>110189029China makes the best small models.America makes better large models but you couldn't run them on affordable hardware even if they were open weights.
>>110189076Indeed, both are terrible.
>>110189089Nothing has ever gotten close to what sora 2 though. The shit that model knew...
>>110187929>>Who are these people? Who the fuck cares about the opinion on this from a bunch of randoms?You can literally fail upwards in America/Brussels as an "Asia Pacific expert". Literally make up random bullshit about China that sounds authoritative, and you'll ping pong from think tank to think tank to NGO (gov funded propaganda) to politics.
>>110187405>American AI companies lost value on the idea that a Chinese model will undercut them>China cuts their own legs off >Their top models are just cheaper distillations of the American modelsGuess I should buy openai and anthropic
>>110189466if you actually use these models chinese models are heavily benchmaxxed cause they are distilled with a fraction of the compute. That said deepseek 1017 has entered the "good enough" tier while being nearly free which is indeed a very big usecase
>>110189482so like in everything else china eventually makes good enough knockoffs for a fraction of the price. Who would have guessed
>>110189482Yeah I use qwen 3.8 27b local. It's good enough for small use but pretty brain dead in long horizon. I could see if they were really topping the charts and not just slashing costs on tokens. Sounds more like internal controls that they don't want local Chinese people using anything but the approved models that will answer on social issues with the accepted answer since releasing globally is objectively better for them financially.
>>110187405Would be terrible if some of those leaked.Reality is, all of those will be instantly hacked and obtained by western glowies.
This was literally the only reason China has a market. Providing uncensored local ai models.Nobody with half a brain trusts Chinese cloud based ai sending all data back home to the CCP for review.But open models that could be isolated and run locally for a fraction of the price of American competitors was enticing.American companies hate it that a product that didn't require billions more in investment but does the same thing competes with them.This stupid move is exactly what US AI companies would want them to do to give US companies a monopoly.The calling home Chinese cloud ai will be banned everywhere as a security threat and American AI will face no competition and citizens will have few options without the mandated censorship.
>>110189940>uncensored not anymore. Kimi 3.1 is more censored than claude and gemini by far. Deepseek is a good deal more censored as well.
>>110189940>Nobody with half a brain trusts Chinese cloud based ai sending all data back home to the CCP for review.but trusts kikes?
>>110189499>Yeah I use qwen 3.8 27b local. It's good enough for small use but pretty brain dead in long horizon.give it memory and automatic handoff at 80% context and it gets pretty good
>>110187405>they the same as anyone else on this planetfell for it again award
>>110190985Thanks I'll try that.
>>110187405Big win for Meta.
>>110189089Not really. They are just more efficient to run.
