Chinese AI app DeepSeek crashes after viral success, limits international access due to alleged mass attacks
Chinese artificial intelligence app DeepSeek has reportedly stopped working due to a technical problem.
The company, little known outside China, has set restrictions so that new registrations can only be made with Chinese phone numbers. In doing so, it effectively blocked registrations from international users. This action appears to be aimed at reducing the number of people interested in accessing and using the app.
A prominent message in the app’s web chat reported that DeepSeek’s online services had been subject to “malicious attacks on a large scale.” However, it did not specify who might be behind those attacks.
Monday’s outages were the latest episode in a series of problems that have occurred in recent days, as DeepSeek achieved viral success. The notion that DeepSeek’s models offer superior performance caused US tech companies to lose $1 trillion in valuation, due to fears that Chinese AI could outperform that developed in Silicon Valley.
DeepSeek was created in 2023, driven by the hype generated by other AI tools developed by companies like ChatGPT. At the time, there was a perception that China was lagging behind the United States in AI research and development.
Since its inception, DeepSeek has released several models that have increased confidence in AI development in China. The company claimed that its most advanced models have similar performance to OpenAI and Meta, while being much cheaper.
AI models from ChatGPT to DeepSeek require advanced chips to power their training. Since 2021, the Biden administration has expanded the scope of bans aimed at preventing these chips from being exported to China and used to train Chinese companies’ AI models.
However, DeepSeek researchers wrote in a paper last month that DeepSeek-V3 used Nvidia’s H800 chips for training and spent less than $6 million.
Although this point was later disputed, the claim that the chips used were less powerful than Nvidia’s more advanced products, which Washington has tried to restrict in China, coupled with relatively low training costs, led American tech executives to doubt the effectiveness of technology export controls.
Little is known about the company behind DeepSeek, a small startup based in Hangzhou, founded in 2023, the same year Baidu unveiled China’s first large-scale AI model.
Since then, numerous Chinese tech companies, both large and small, have developed their own AI models. However, DeepSeek became the first to receive praise from the American tech industry for matching, and even surpassing, the performance of the most advanced U.S. models.
