Chinese AI model DeepSeek faces cyber attacks after disrupting global market (includes video story)

A new Chinese artificial intelligence model, DeepSeek, has quickly gained attention, surpassing its U.S. counterparts in the Apple App Store. The AI chatbot, which directly competes with advanced models from American tech giants like OpenAI, Meta and Google, also sparked concerns among owners of AI related stocks, some of which experienced a significant drop on Monday.

The platform, known for its affordability and rapid development, was hit with a large-scale cyber attack, prompting the temporary suspension of new user registrations outside of China. The incident has highlighted DeepSeek’s recent success as well as its vulnerabilities.

DeepSeek, which uses a model called R1, has been hailed as a major advancement in artificial intelligence. The Chinese startup, just over a year old, caused the NASDAQ to drop by 3% on Monday as investors scrambled. 

The company claims that it spent just $5.6 million to develop the model, a fraction of the cost associated with other major AI systems. According to DeepSeek, its R1 model was trained in under two months, far less than the many months required by other AI platforms. 

“The magic here is that instead of taking months to train and hundreds of millions of dollars, it took under two months to train and cost, according to the company, under $6 million to train,” explained Shelly Palmer, a professor of advanced media at Syracuse University.

Its use of less advanced chips is a strategy driven by U.S. export restrictions on China. While other companies have been limited by the lack of access to cutting-edge technology, DeepSeek has used this constraint to its advantage, finding more affordable alternatives that still deliver impressive results.

“OpenAI has been in a monopolistic position,” said Mark Finlayson, an associate professor of computer science. “Optimally, no one’s had a model that worked as well as they did, partly because of the data and the first-mover advantage. They were always a little bit ahead.”

Experts warn that its relatively low cost and unorthodox development strategy could expose the model to significant security risks. According to cybersecurity firm Keller, DeepSeek’s R1 model has proven to be more vulnerable to cyber attacks than its competitors. 

The attack, which began in early January, reached its peak this week with a massive Distributed Denial-of-Service (DDoS) assault on the platform’s API and web chat services. The DDoS attack, which bombards a service with traffic to overwhelm its capacity, is believed to have originated from multiple countries, including the U.S., Singapore, and the Netherlands.

While DeepSeek has yet to fully recover from the attack, the company has pointed to U.S. IP addresses as potential sources of the breach. The U.S. government has not yet commented on the situation.

Samantha Gutierrez is a student journalist majoring in Digital Communication and Media. As a bilingual reporter, she enjoys covering stories about her diverse Miami community. With a strong passion for broadcasting, she aspires to pursue a career as a multimedia journalist in South Florida.