Victim of a cyberattack, DeepSeek limits registrations

© Poetra.RH/Shutterstock.com

For the past few days, a new chatbot called DeepSeek has been shaking up American tech. Developed in China, it is based on a language model that rivals o1, OpenAI’s model specialized in reasoning, while its training required only a fraction of the resources used by OpenAI. The release of DeepSeek caused a drop of more than 15% in Nvidia’s stock price. And as I write this article, DeepSeek is number 1 in the free apps ranking on the American App Store.

But if you want to try this new AI now, you will have to wait. It is possible that, for the moment, you will not be able to register on the application. Indeed, as CNBC reports, DeepSeek has indicated that it is temporarily limiting new signups “due to large-scale malware attacks” targeting its services. However, those already signed up will be able to continue using DeepSeek’s AI.

At the moment, we don't have all the details of this cyberattack and we don't know when registrations will resume. But in any case, DeepSeek will probably create a buzz in the coming weeks. Indeed, in addition to the fact that it was developed with paltry means compared to the means of American AI giants, DeepSeek's artificial intelligence was also developed despite American restrictions on chip exports.

Open source highlighted ?

In a publication on the social network Threads, Yann LeCun, chief AI scientist at Meta, explained in a publication that this is above all a victory for open source over closed models. “To people who see DeepSeek's performance and think: “China is overtaking the United States in the field of AI, you are misreading it. The correct reading is: “Open source models are outperforming proprietary models,” he wrote.