JAKARTA DeepSek's security is still a concern to date, even this chatbot is still being blocked in several countries. Now, the latest findings say that DeepSekcan develop malware.

This was revealed by Tenable Research, a security solutions provider based on Artificial Intelligence (AI), on Friday, March 14. The generative AI model should have built-in protection to prevent deviant things its users do.

However, after being researched by Tenable, DeepSek R1 can actually be tricked into creating malware. In other words, cybercriminals who are lazy to make their own malware can be helped by DeepSek. This is a very large security risk and needs to be considered.

In fact, the results DeepSek provides still need to be improved manually in order to function effectively. If these irresponsible users do not have the knowledge to create malware, the code DeepSek gave will not be developed.

However, it is different if the person is very good at it. With only a few improvements, coding from DeepSek can be a real malware. This raises concerns because DeepSek is easily penetrated and can be tricked to help crime.

Nick Miles, Tenable Research Technical Staff, said that DeepSek originally worked well, namely refusing to make a keylogger. However, after submitting the same request as other approaches, DeepSek immediately opened the limits.

"However, by rearranging the request as a 'education exercise' and implementing the generalized jailbreaking method, we quickly overcome the limits," Nick said in a broadcast received by VOI.

From the test results, it can be concluded that DeepSeek is capable of generating log-encrypting keyloggers and keeping them secretly on devices and producing executable ransomware that can encrypt files.

Tenable research highlights the urgent need for responsible AI development, explains Nick. "Organizations, policymakers, and security experts must work together to ensure that this sophisticated device does not drive cybercrime," Miles said.


The English, Chinese, Japanese, Arabic, and French versions are automatically generated by the AI. So there may still be inaccuracies in translating, please always see Indonesian as our main language. (system supported by DigitalSiber.id)