ChatGPT killer or privacy risk: What is DeepSeek and why is everyone talking about it?

It’s nearly impossible to escape the online hype surrounding DeepSeek, a relatively new and unknown AI chatbot, right now. In just a few days, it has not only dethroned ChatGPT’s dominance in benchmarks but also become the most downloaded app on iOS and Android. What’s even more impressive is that the AI was developed by a small Chinese startup with a tiny budget and relatively outdated hardware.

So what exactly is DeepSeek, why has it garnered so much attention, and should you consider using it over ChatGPT or Gemini? I dug into the Internet’s new darling AI and took it for a test-drive to find out.

What is DeepSeek?

On the surface, DeepSeek is simply an AI chatbot that you can use for free. In these terms, it’s functionally the same as OpenAI’s ChatGPT or Google’s Gemini. However, unlike its competitors, DeepSeek doesn’t impose any limits or charge a monthly subscription to use its most capable models. It’s also open source, meaning that you could download and run DeepSeek locally — as long as you have hardware that’s powerful enough.

DeepSeek’s latest models don’t just come close to matching the competition; they often surpass it in several areas. The latest DeepSeek-V3 model scores better in several coding, math, and Chinese language benchmarks than OpenAI’s GPT-4o and Anthropic’s Claude-3.5. The latter were regarded as the gold standard in AI performance until DeepSeek dethroned them practically overnight.

What’s so special about DeepSeek?

DeepSeek has been around for some time now, but it wasn’t until January 20, 2025 that it began gaining mainstream popularity. That day saw the release of DeepSeek-R1, the company’s first AI model capable of reasoning like a human.

Even as a fairly regular AI user, I have to admit that the DeepSeek-R1 model seems magical in how it seems to think before responding to my prompts (pictured above). While we’ve already had AI capable of reasoning since OpenAI released GPT-o1 in September 2024, it has remained out of reach for most people. That remains true even today — you’ll need to fork over $20 per month for a ChatGPT Plus subscription to access the GPT-o1 reasoning model.

For developers looking to integrate AI models into their own apps, DeepSeek is about 20 to 30 times cheaper as compared to ChatGPT’s underlying model. All of these factors combined make DeepSeek a formidable contender in the AI race, even though it appeared out of virtually nowhere.

DeepSeek vs ChatGPT and Gemini: Will I switch?

While DeepSeek technically outperforms other AI chatbots, I don’t think it’s the breath of fresh air that the internet is currently making it out to be. For one, the model is still just as prone to hallucinations as any of its rivals and I’ve found that it will absolutely invent facts out of thin air when asked about something it’s unfamiliar with. This makes its usefulness situational at best — and not that much different than ChatGPT after all.

Even DeepSeek-R1, the model capable of human-like reasoning, only makes sense in very limited use-cases. Unless I’m writing complex code or solving math problems on the regular, I won’t get any better results from the reasoning model than the standard DeepSeek-V3 model. The latter doesn’t let me observe the AI debate itself but it’s significantly faster and nearly as adept at answering questions and other general-purpose tasks.

I can’t say there are many incentives to make the switch to DeepSeek right now, even as a regular ChatGPT and Gemini user. The latter have a much more polished ecosystem, with features like vision and two-way voice chat a la Gemini Live that I use much more frequently. DeepSeek only supports text-based conversations for now, although that will likely change sooner rather than later.

Then there’s the elephant in the room: privacy and censorship. While I have no qualms with DeepSeek’s Chinese origins, that’s only because I don’t tend to use AI for any tasks that require discretion or secrecy. Given that OpenAI has to contend with data privacy regulations in Europe and North America, you can at least change a setting to prevent your chats from being used to train future models. The same cannot be said for DeepSeek, unless you host it yourself.

If you use the online version of DeepSeek, you’ll also have to deal with censorship. The chatbot currently avoids talking about topics sensitive to Chinese sentiments, but it could begin promoting biased information in the future. So if you’re looking for the most honest chatbot out there, DeepSeek does not exactly fit the bill.

Who created DeepSeek and why does it matter?

DeepSeek is owned by Chinese entrepreneur Liang Wenfeng, who also created a hedge fund named High-Flyer. The startup’s outstanding performance would have gone largely unnoticed outside of the AI world if it weren’t for its Chinese origins and almost shoestring budget. DeepSeek has managed to dethrone billion-dollar ventures like OpenAI while also proving that larger investments don’t always result in better outcomes.

China has historically lagged behind the West in the AI race, largely due to the U.S. government imposing strict export controls on American companies like Nvidia starting in 2022. These controls banned the sale of advanced AI training and processing hardware to Chinese companies. Moreover, without the support of tech giants like Microsoft and Google to pour billions of dollars into AI research and development, it seemed unlikely that China would ever catch up.

Yet, we now know that a lean Chinese startup managed to develop a highly capable AI model with allegedly just $6 million in computing power — a fraction of the budget used by OpenAI or Google. DeepSeek achieved this feat using older NVIDIA H800 GPUs that it managed to acquire in spite of the US’ export controls. The chatbot also uses homegrown Huawei-made chips to generate responses, further proving that China doesn’t need American hardware to compete in the AI race.

Why did DeepSeek affect US tech stock prices?

DeepSeek’s success has rattled both the U.S. government and American tech giants, but likely only in the short term. NVIDIA’s valuation dropped by $600 billion while the tech stock-heavy NASDAQ 100 index dropped 4% in a single trading session. However, most of them reversed course the very next day, signalling that the market is still optimistic about the future of AI. A rising tide lifts all ships, after all.

Why do governments want to ban DeepSeek?

After TikTok, DeepSeek might become the latest app to draw the ire of the US government and even face a total ban over national security concerns. While the chatbot does not collect data directly from your phone, we don’t know how the company collects and stores its users’ chats on its servers. It’s entirely possible that the Chinese government can gain access to all user data, which would have grave consequences for national security. Given that, it’s not surprising that we’ve seen the Australian and American governments urge caution when using the chatbot.

Currently, DeepSeek’s main advantage is in achieving cutting-edge performance with far less computing power than the competition. Luckily, that doesn’t have much impact on the way you and I use AI today, so a ban would have little impact in the short term.