Deepseek, Chinese launch AI, which has recently captured most of artificial intelligence (AI), said it limits the registration to the service, citing malicious attacks.
“Due to large-scale malicious attacks on Deepseek services, we temporarily limit the registration to provide permanent service,” the company – Note On the incident report page. “Existing users can log as usual. Thank you for understanding and support.”
Users try specify The account shows a similar message that states that “registration can be busy” and that they should wait and try again.
“Deepseek’s popularity is not a big surprise that they are aimed at malicious traffic,” said Eric Kron, who a security defender in Knowbe4, said in a statement shared with Hacker News.
“Such attacks can be a way of extorting an organization, promising to stop attacks and restore the fee, it may be competitive organizations that seek to negatively affect competition, or even there may be people who invest in a competitive organization and want to protect their investments by raising Competition. “
Deepseek, founded in 2023, is a Chinese beginning that “devoted to the creation of AGI (artificial general intelligence)”, according to description On his hug page.
Company has become a conversation in the world AI, with an app iOS chatbot achievement This week at the top of Apple’s free applications in the US, the Openai CHATPT is.
The company has released a number of reasoning and language models of the mixture experts under the MIT license, which claims US sanctions This prohibits the sale of advanced AI chinese companies.
“During the pre-training phase, Deepseek-V3 training on each trillion token requires only 180k H800 GPU hours, that is 3.7 days on our cluster with 2048 H800 GPU, company company company – Note In the study.
“Thus, our previous workout stage ends in less than two months and costs 2664k hours GPU. Its complete training.
This is said to have been found the platform Answers Censorship to sensitive topics such as Tiananmen Square, Taiwan and Uighur treatment in China.
His privacy policy too notes User personal information – including devices and networking information, usage models and payment data – posted in “Safe servers located in the People’s Republic of China” Fresh problems For Washington Among the prohibition of the ciking.
‘We live in a temporary scale where the company is not in the US – Note Jim Fan, Senior Research Manager and Leading AI (Gear Lab) in Nvidia.
CEO Openai Sam Altman called The Deepseek’s R1 reflection model is “impressive” and that it is “legitimately encouraged to have a new competitor.”
