Deepseek launched its new language model, R1last 20. In just one week, the chatbot has become the sensation of the industry, standing as The number one app in downloads in the App Store in the United States and causing an earthquake in the stock market capitalization of many technological companies. Nvidiathe most advanced hardware manufacturer for the training of generative artificial intelligences, took the worst part, losing in a single day around 17% of its stock market value, about 600,000 million dollars.
There are two reasons behind Depseek’s success. The first, the new language model offers results comparable to those of O1the most advanced model that uses Chatgpt And it is paying, but No need for a subscription. Second, The cost. R1 has been trained by a fraction of what other language models cost. Deepseek did not reach 6 million dollars, while Openai spent 100 with GPT-4.
With everything good that this may sound, we must remember that nothing is free on the Internet and that user data has great value for companies. In the case of Deepseek, this data is stored on servers in China. ‘We store the information we collect in safe servers located in the People’s Republic of China’says the company in its privacy policy.
Tiktok’s precedent
This, as you know Bytedanceit can be problematic. The Chinese company that owns Tiktok faces to the prohibition in the United States due to the concerns generated about national safety and privacy of users data. Chinese companies are obliged to facilitate the data that the Government of Xi Jinping asks, so it is a risk to the privacy of users who store in China. This was the case of Tiktok in the United States for years, until Bytedance reached an agreement with Oracle In 2020 to store the data of US users in their cloud infrastructure, but it has not been enough to placate the government that He wants the Chinese company to sell its operations in the United States so that Tiktok is not prohibited.
With these precedents, and the addition of the disruption that Depseek is generating in the technological companies of Silicon Valley, The Chinese company is likely to face the same problem sooner rather than later.
What data is Deepseek and how to prevent them from sending them to servers in China
According to Wired, the data that Deepseek collects from users are framed in 3 areas: the information that the user shares with the chatbot, which he collects automatically and the one he obtains from other sources. This covers from the messages, audios and files that the user introduces to technical aspects such as the device that is used or the IP address, through others such as the information associated with the Google account with which session is logged in or the patterns to the typing of the user.
Deepseek offers the option to erase the chats with AI, but assuming that this means that they are really eliminated depends on the confidence that the Chinese company is wanted to give. Now, Yes there is a safe way to use the app and prevent the data from ending in China.
When we talk about Deepseek, we are doing it two things. On the one hand, the Chatbot that can be used via web or through the app; on the other, of the open source language model which supports it and anyone can download and execute at home, with user control over privacy.
This is what allows Perplexity.ai offer Deepseek R1 among the models that the user can choose with their chatbot. The company ensures that ‘Deepseek in Perplexity is housed in data centers in the United States and the European Union: Your data never leave western servers. The open source model is accommodated completely independent of China. Your privacy and the safety of your data are our priority. ‘
However, he has his counterpart and is that he loses one of the main attractions. To use Depseek R1 with perplexity, it is necessary Subscribe to the Payment Plan for the search engine based on AI; $ 20 per month that also give access to advanced models such as Openai O1 or Claude 3.5 Sonnet, among others.