The Chinese generative AI platform DeepSeek has recently gained significant traction, but this popularity has come with heightened scrutiny. Security researchers discovered that one of DeepSeek’s critical databases was left exposed online, leading to the leakage of over a million records, including sensitive user data and API keys.
The cloud security firm Wiz reported that DeepSeek was unresponsive when researchers attempted to alert them about the exposure. They faced challenges in communicating their discovery and instead sent messages to various DeepSeek-related contacts they could find. Shortly after their outreach, the exposed database was secured, leaving uncertainties about whether any unauthorized access had taken place.
Wiz’s Chief Technology Officer, Ami Luttwak, highlighted the severity of the oversight, emphasizing that this type of exposure indicates a lack of maturity in DeepSeek’s security practices. The vulnerability was notably easy to find, suggesting that exposed databases remain a recurring issue in the tech community.
The leaked database appeared to be a ClickHouse database, traditionally used for server analytics. It contained logs of user interactions, prompts, and authentication tokens, primarily in Chinese, though there may have been entries in other languages as well. Wiz researchers speculated that hackers could have exploited the access point to infiltrate deeper into DeepSeek’s systems.
Security expert Jeremiah Fowler pointed out the shocking nature of such a lapse in security for an AI platform. The exposed operational data could significantly risk the company’s integrity and user privacy. The findings reflect broader security vulnerabilities present in cloud-hosted databases, even in advanced technological frameworks.
Despite the ongoing rise in users for DeepSeek, which climbed to the top of app store rankings, the fallout from these revelations has raised alarms among U.S. AI companies and regulators. There have been concerns about data privacy, censorship practices, and implications of Chinese ownership on national security.
Regulatory bodies, including Italy’s data protection authority, have begun probing DeepSeek regarding the sources of its training data and the handling of personal information. Concurrently, reports indicated that the U.S. Navy advised personnel against using DeepSeek due to potential security and ethical risks.
As the AI landscape evolves rapidly, the recent exposure serves as a stark reminder of the persistent security challenges that continue to plague even the most innovative technologies in the industry.
