Microsoft has taken strong action against DeepSeek, the Chinese generative AI that presents itself as an alternative to ChatGPT. During a hearing before the United States Senate, Brad Smith, Microsoft's corporate vice president and president, revealed that the company's employees were not allowed to use the chatbot. The executive pointed out that DeepSeek is also not available on the company's app store, the Microsoft Store.
To justify the ban, Smith points to data risks. Unsurprisingly, Microsoft fears that information shared with DeepSeek could end up stored on servers in China. If Microsoft employees were to use Chinese AI, the company's data, including confidential information, could end up in the hands of DeepSeek-Vision, a Shanghai-based startup. In addition, Microsoft fears that the AI's responses could be influenced by propaganda from the Chinese Communist Party, which is a major player in China. Unsurprisingly, DeepSeek censors certain topics.
Since its appearance last January, DeepSeek has raised numerous concerns regarding confidentiality and privacy, especially in the United States. A U.S. senator even raised the possibility of criminalizing DeepSeek's use. The Chinese AI's terms of use clearly state that it collects information. DeepSeek openly acknowledges the fact that it collects the personal data of users who interact with it.
Double standards?
Despite DeepSeek's potential dangers, Microsoft has not hesitated to offer the DeepSeek-R1 model to Azure customers. Developers can use the group's computing power to run the AI model. There is obviously a nuance between using the model and using the DeepSeek application. Since the model is open source, it can run without problem on servers located in the United States, under Microsoft's control. In this scenario, the data is unlikely to end up in China.
However, it is theoretically still possible that Chinese propaganda could influence the responses of generative AI, even if the servers running the model are not located in China. Before the Senate, Brad Smith clarified that Microsoft managed to access the language model to correct it. Adjustments have been made to avoid "harmful side effects."
As a reminder, DeepSeek is likely to be banned outright in the United States. The Trump administration is considering banning Chinese artificial intelligence from government devices and banning the app from the Play Store and App Store. Other nations have already taken restrictive measures against Chinese AI. This is the case in South Korea, which has banned DeepSeek from all app stores, or Australia, which has banned DeepSeek from all government systems.
Source: TechCrunch
0 Comments