The DeepSeek AI innovation is all the tech world is thinking about right now. It’s a great example of Chinese innovation despite the sanctions that prevent most AI firms from competing against ChatGPT, Gemini, Claude, Meta AI, and other AI chatbots from the US and Western world. It’s also a great example of China retaliating against the US in the early years of the AI race. After being hit by sanctions, China responded in an unusual way.
DeepSeek excelled at software optimizations to deliver a reasoning AI model that’s as good as ChatGPT o1 at a fraction of the cost. The release of DeepSeek R1 wiped out nearly $1 trillion from the US stock market, with NVIDIA among the top losers. At the same time, DeepSeek became the most downloaded iPhone app in the US App Store.
But as exciting as it might be to use DeepSeek right now, you should remember an important problem. All your DeepSeek data is sent to China, including the chats you have with the AI. You only get great privacy with DeepSeek if you run the open-source model locally on a computer rather than the iPhone or Android apps.
Some of the massive amounts of data DeepSeek gobbles up will be used to train DeepSeek R1 successors. Meanwhile, ChatGPT doesn’t get the same privillege, as it’s not exposed to the massive Chinese market.
DeepSeek privacy and the implications of having an AI model trained in China that could possibly mislead US users could be enough for some US government officials to consider banning DeepSeek AI and other similar products from China.
This isn’t just about the TikTok precedent here. It’s about reciprocity, considering that China doesn’t allow products like ChatGPT, Gemini, Claude, and Meta AI. Apple Intelligence is also unavailable in China, as Apple has to partner with a local AI provider to release it. By the way, TikTok is also unavailable in China.
DeepSeek is such a resounding success in China because the US and other international markets do not have the same censorship rules in place as China. DeepSeek AI is available on the iPhone and Android because international developers can compete, whether it’s Chinese AI firms or any other companies. As long as developers respect Apple and Google’s rules, they can access the App Store and Google Play.
You can run DeepSeek on the web like ChatGPT because the internet isn’t censored in the US like it is in China. That’s also why you can install an open-source version of DeepSeek on your computer.
All of the above would apply to any app. It’s why TikTok became such an international success. Good software can compete and win.
But the same doesn’t apply to China. When ChatGPT took the world by storm in late 2022, but wouldn’t have produced a similar effect in China to what DeepSeek just did in the West. None of the AI innovations that followed, whether Google Gemini or Meta AI, would be available to regular consumers in China.
Some can always try to circumvent the Great Firewall of China to access US AI programs. But it’s a hassle and might not be worth pursuing. Also, OpenAI blocked access to ChatGPT in China last summer, so it’s even more difficult for anyone to use it.
Meanwhile, AI firms like DeepSeek certainly had access to DeepSeek and other AI programs. Meta’s Llama models are open-source. Some early versions of DeepSeek identified themselves as ChatGPT in chats, suggesting DeepSeek might have used US AI models to train DeepSeek. That’s just speculation based on the observed behavior of DeepSeek.
Even the fact that DeepSeek users in the US and elsewhere could test the early versions of DeepSeek shows how easy it is for Chinese tech firms to deploy AI and non-AI software in the US compared to the other way around.
As a longtime ChatGPT user who isn’t likely to quit using AI chatbots anytime soon, I see the DeepSeek R1 release as a great development. The market will recover, and the DeepSeek solutions will surely be implemented by more AI companies, including the likes of OpenAI. This will lead to more innovation and more steps towards AGI (artificial general intelligence) and artificial super intelligence (ASI). In turn, I’ll get better, more reliable chatbot experiences for cheaper prices. That’s the DeepSeek revolution.
The problem with DeepSeek is that I would still worry about using it. It seems like it’s just like TikTok, but it can be worse. US lawmakers didn’t just worry about TikTok sending user data to China; they singled out the TikTok algorithm and its ability to manipulate users. Build a similar algorithm into an AI model you then offer for free, and the risks are even bigger.
I’m not saying that DeepSeek AI will do that, but it could open the doors to US customers accepting more AI models from China. And those will come. They might be just as cheap and fast as R1. What if one or a few of them come with specific instructions to subtly sway public opinion in favor of China or Chinese interests?
Also, look at how powerful TikTok became in the US and the world at large as it grew more popular with users. President Trump is the best example of that. He wanted a ban in his first term to now be one of the app’s defenders. TikTok played a role in having him reelected.
While I think strong competition in the AI space could benefit the world greatly in the coming years rather than doom it, I can’t see how DeepSeek AI can thrive in the US when ChatGPT isn’t available on a similar scale in China.
