\n<\/p>\n
As Chinese open-weight AI models grow in capability and popularity<\/a>, arguments about what should be done about them<\/a> have once again reached a fever pitch.<\/p>\n There\u2019s talk<\/a> that the Trump administration might try to ban them (though it hasn\u2019t yet acted<\/a> on the idea). Meanwhile, proprietary model makers, particularly OpenAI and Anthropic, appear<\/a> increasingly concerned about them.<\/p>\n Open-weight models such as Moonshot AI\u2019s Kimi K3 or Alibaba\u2019s Qwen offer inference at a fraction of the token cost of closed-source models from these large U.S. labs. The fear is that they also pose some sort of threat. Certainly, they threaten the profit margins of the large proprietary AI labs. <\/p>\n But should enterprises running these models in their own data centers succumb to the fear that they could be a vector for Chinese hackers? <\/p>\n No, says Lucas Atkins, the CTO of Arcee<\/a>, which is building open models to give U.S. companies a homegrown alternative to Chinese models.<\/a><\/p>\n If any startup would benefit from a ban on Chinese models, Arcee would. But Atkins says China\u2019s open models are no more dangerous than any other open-source software a company may use. In fact, he says, they even offer benefits even to his own company. <\/p>\n \u201cA lot of people view this as similar to a Chinese software program. Like, it was coded with these x, y, z intentions\u201d that a bad actor could simply command, he said.<\/p>\n \u201cThat is fundamentally not how these models are trained. There is really not any way for an Arcee, or an Alibaba, to make a model, have someone run it in their own environment and for us have any access to it whatsoever,\u201d he explained.<\/p>\n While most of these models are what\u2019s known as \u201copen weight,\u201d and are not really fully open-source software, the source code (the part that will actually run on servers), if it is downloaded from open-source sites like Hugging Face, is similarly largely visible and reviewable. (What isn\u2019t available is the methods and data used to train the models.) <\/p>\n Large organizations should put any model core through their security testing and inspection processes, and they will also often post-train the models for their specific uses, and can examine areas like bias, toxicity, hallucinations, sensitivity to certain topics. So they work with, optimize, and understand the models before people start sending them prompts.<\/p>\n Could a model that is used for coding somehow throw malicious backdoors into the code it writes? Again, while that\u2019s theoretically possible, it would require acrobatic feats to accomplish.<\/p>\n \u201cThere\u2019s no reason that a sophisticated enough actor couldn\u2019t train a model to be a completely amazing coding model in every circumstance, but when presented with a certain type of code base \u2026 some hidden training would kick in,\u201d Atkins, who spends his days training models, postulated. But he adds: \u201cI don\u2019t know how you would do this.\u201d<\/p>\n Because large language models are by nature creative, the odds are slim of getting a contemporary model to spit out malware in response to a preplanned perfect storm of context and prompt. Even slimmer are the chances that any enterprise would then use that code.<\/p>\n Could it happen in the future? That\u2019s anyone\u2019s guess. But enterprises are also building their AI apps to be model-agnostic and to use multiple models. So even if Chinese models are the best for the price today, enterprises won\u2019t be locked into using them forever.<\/p>\n \u201cI think instead of the conversation being about how to ban Chinese models, it should be about how do we foster a good, open ecosystem here in the U.S.,\u201d Atkins says.<\/p>\n Arcee also gains advantages from Chinese models. Because they are open, the startup \u201cbenefits from those models being good because we can learn what they did. We can build on top of them. Then they can learn what we do,\u201d he says. \u201cWe have tremendous respect for the people building those models, the individual researchers.\u201d<\/p>\n Ultimately, the way to compete with Chinese models \u201cis to release a model that is better,\u201d says Atkins. \u201cWe need to give them something to talk about.\u201d<\/p>\n<\/div>\n When you purchase through links in our articles, we may earn a small commission<\/a>. This doesn\u2019t affect our editorial independence.<\/em><\/p>\n