DeepSeek Grapples with Major Outages and Large-Scale Cyber Attacks

DeepSeek, a leading Chinese open-source AI company, experienced significant disruptions in its services this morning due to "malicious attacks," highlighting its impactful role in the industry.

Significant Disruptions

This morning, AI giant DeepSeek experienced service interruptions affecting its APIs and web chat services. This Chinese startup, which currently tops Apple’s App Store free apps chart, suffered major performance setbacks, much to the frustration of its users.

A Major Player in AI

Founded in 2023 and backed by the hedge fund High-Flyer, DeepSeek, under the leadership of Liang Wenfeng, cofounder of High-Flyer, has quickly established itself as a significant force in the AI landscape. Based in Hangzhou, Zhejiang, the company is dedicated to advancing general AI through open-source research and development, aiming to make AI technology widely accessible for commercial and academic applications.

Temporary Restrictions

In response to “large-scale malicious attacks” on its online services, DeepSeek has temporarily limited new registrations to individuals with a China-based phone number. Existing users are not affected and can continue to access their accounts without disruption.

Commitment to Open-Source

Unlike many AI companies that offer subscription-based models, DeepSeek fully embraces open-source. Its models are available under the MIT license, allowing unrestricted commercial and academic use. This commitment to openness has significantly contributed to its growing popularity.

The company has introduced several groundbreaking models, including DeepSeek-V3, which features 671 billion parameters and was trained on a dataset of 14.8 trillion tokens in about 55 days at a cost of approximately $5.58 million. This model outperforms others such as Llama 3.1 and Qwen 2.5 in terms of capabilities, while matching those of GPT-4o and Claude 3.5 Sonnet. Its architecture incorporates a mix of experts with a Multi-Head Latent Attention Transformer, featuring 256 routed experts and one shared expert, activating 37 billion parameters per token.

By demonstrating that it is possible to develop high-performing AI models at significantly lower costs and resources, DeepSeek has prompted a reevaluation of massive investments in AI infrastructure by major tech companies.

Awaiting Further Details

The company has not provided specific details about the nature of the malicious attacks or the expected duration of the registration limitations. This situation highlights the security and service availability challenges faced by rapidly growing AI platforms amid increasing cyber threats.