DeepSek Is Still At Risk Even Though It's Been Blocked, Expert: It's Not That Easy

JAKARTA - Several countries including the United States (US), Italy, and Australia, have banned access to DeepSeek, a large language model (LLM) developed by China, on government devices.

The ban arose due to concerns over national security and the potential for sensitive data sharing of these users to the Chinese government.

Seeing DeepSek blocking in several countries, a senior staff research engineer at Tenable cybersecurity firm Satnam Narang commented that this ban is easier to implement on mobile apps or websites.

However, because DeepSek is an open AI model, this blocking is certainly much more difficult than blocking access to an application or website.

"Because DeepSeek LLM itself is open-source, blocking access to it is not that easy, because it can be run locally on the device, can be run through the cloud environment, and installed through various applications or frameworks such as the ollama," Narang said in a written statement received by VOI, quoted on Sunday, February 9.

Although it seems that accessing and installing DeepSek looks more difficult than in countries that don't ban it, it believes users can still do it.

He also added that although local LLMs do not require connections to external servers, reducing the risk of data leakage, other challenges remain.

DeepSek websites and applications can indeed be blocked on government devices, but according to him, DeepSek can still be accessed through private devices, which have the potential to be used to share sensitive information.

Another worrying aspect of DeepSek is not only exposure to sensitive information, but also the lack of security features on the model, as it can be used for harmful or malicious purposes, he stressed.