Developing an AI model from scratch isn’t easy, and doing so doesn’t guarantee it will be good either. We’ve seen Apple attempt this with Apple Intelligence to disastrous results before throwing in the towel and working out a deal with Google. But not everyone can afford to license other AI models for $1 billion a year, so some have resorted to more dubious methods. This is something that Chinese AI companies are allegedly doing, so much so that the US is threatening sanctions against China over AI IP theft.
US threatens China with sanctions over alleged IP theft
According to Treasury Secretary Scott Bessent, the US government will look into open-source AI models from China for signs of IP theft. If the US determines that Chinese companies engaged in IP theft, sanctions could be on the table.
Speaking on Fox Business Tuesday, Bessent said, “We’ve seen a lot of talk about open source models coming and threatening the large language models in the U.S. This administration supports open source models, but what we do not support is IP theft. If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft.”
So, how do these thefts take place? Basically, some unscrupulous AI companies leverage thousands of fraudulent accounts to flood AI models like ChatGPT or Claude. They then capture the output generated by these models and use them to train a smaller, cheaper model. This is a similar process Apple went through with Gemini, where it distilled the training from Gemini and used it in Apple Intelligence. However, in Apple’s case, it was a licensed deal.
Not the first time
We’ve heard about this before. Back in 2025, when China’s DeepSeek hit the market, many were surprised at how quickly it was deployed and yet how powerful and capable it was. Early testing found that it was capable of matching or even surpassing other established models.
Eventually, security researchers at Microsoft claimed to have found evidence that suggested that DeepSeek might have trained itself on OpenAI’s GPT models. To date, the company behind DeepSeek has yet to comment on the allegations. However, subsequent updates to the model removed references made to GPT or DALL-E.
The post US Threatens China with New Sanctions Over AI Model Theft as Tech Tensions Escalate appeared first on Android Headlines.
