Chinese startup DeepSeek has taken the world by storm in a matter of days with a ChatGPT o1 reasoning model rival that’s reportedly much cheaper to train and doesn’t need as much compute and energy. The release of the DeepSeek R1 model and research contributed to tanking the AI market, as some investors may fear the AI bubble could burst.
Chinese AI startups do not have access to the latest chips, including NVIDIA’s GPUs that the likes of OpenAI, Google, and others use. Yet DeepSeek was able to achieve the same performance as ChatGPT o1 by changing the way you train a reasoning model to, well, reason.
Unsurprisingly, DeepSeek rose to the top of the App Store, surpassing ChatGPT in the process. People flocked to test the service regardless of the obvious downsides of using a product like DeepSeek.
As of this writing, DeepSeek continues to be the top iPhone app in the App Store. However, new users might have problems accessing the service. DeepSeek is limiting registrations, blaming mysterious cyberttacks.
The cyberattacks
This is the message you’ll see more than once on the status page the company has up and running:
Due to large-scale malicious attacks on DeepSeek’s services, we are temporarily limiting registrations to ensure continued service. Existing users can log in as usual. Thanks for your understanding and support.
The message was first posted around noon Eastern Time on Monday. The same update appeared at the top of the page early on Tuesday. As you can see in the image above, DeepSeek posted the same message on the DeepSeek login page.
The “malicious attacks” appear to be an ongoing issue, but it’s unclear what DeepSeek might be referring to. It wouldn’t be surprising to see hackers and even nation-states go after a new viral service like DeepSeek.
It’s important to note that DeepSeek is only restricting new registrations. Those users who have made an account with the company can log in and use the chatbot.
It’s unclear whether these malicious attacks have targeted user data in any way.
What if there’s another explanation?
I can’t help but wonder whether DeepSeek has the infrastructure needed to accommodate such a massive influx of traffic. Millions of iPhone users might have installed the mobile app to try the service. Others must be accessing the AI via the web. More than a million Android users also downloaded the app.
If DeepSeek can’t meet the surging demand, downtimes, and slow responses are to be expected.
The DeepSeek R1 training breakthrough is what makes me wonder whether the company can take in all the traffic. It’s one thing to be more efficient than OpenAI at training a ChatGPT rival and another to offer dependable AI service to a growing list of users.
No matter how efficient the training and how cheap the access to its models might be, DeepSeek would need to scale up infrastructure to meet the rising demand. However, due to the ban, DeepSeek’s AI experience and access may be impacted if it doesn’t have access to the latest innovations in AI hardware.
I’ll also speculate that if DeepSeek is experiencing a surge in traffic, blaming registration issues on malicious attacks plays in its favor. China just delivered a big blow to the US with DeepSeek R1, tanking the market by showing that AI training doesn’t necessarily need massive resources.
DeepSeek acknowledging that it can’t accommodate new users would imply that the chip sanctions are working. That, training breakthroughs aside, it can’t really compete with OpenAI and Google, and that compute continues to be a key resource for AI development. An ongoing attack sounds better.
While I haven’t used DeepSeek myself so I can’t judge how crowded the service might be, there is one telling review in the Play Store that indicates the company can’t match peak traffic. A user gave the service three stars, noting significant latency issues:
Works great when it’s not overloaded. However, it seems that when I’m attempting to use the DeepThink R1 and Internet search options in communicating with it, the AI experiences significant latency issues. It seems this AI is super popular, and more people are using than the servers can handle. Hopefully, they upgrade their hardware to address this.
I’m saying all this because we’re in the early days of an AI arms race between the US and China, with the former being in the lead. That’s something to always have in mind, whether it’s trying to register for DeepSeek, considering the ChatGPT alternative for your own apps and services, or just hearing that DeepSeek might be under cyberattacks.
DeepSeek AI might be open-source, but there is a lack of transparency here. This is an AI coming from China, which means it’s been trained with censorship in place. Just try asking DeepSeek about events in China the government doesn’t want to remember or talk about. With that in mind, it’s easy to see why DeepSeek and/or the government might want to temporarily hide the fact the company can’t deal with a massive influx of traffic.
That doesn’t mean that DeepSeek can’t be the subject of cyberattacks. Both things can be true at the same time.
Speaking about transparency, DeepSeek should also inform users whether these hacks have targeted existing accounts, stored user data, AI chats, source code, and so on.
