The new DeepSeek AI assistant has attracted a lot of attention in recent days. Kaspersky experts have also detected scam activity related to it.
Due to high numbers of new users and an alleged cyberattack on DeepSeek, there are glitches in the registration process on DeepSeek’s app and website – many registrations cannot come through.
This situation can be used by cybercriminals to steal the credentials of users through fake DeepSeek web pages.
Through such fake registration pages, attackers can collect users’ emails and passwords. These can be exploited to access users’ accounts – on DeepSeek or in other services (if the password is the same for multiple accounts).
There were also several new crypto tokens based on the DeepSeek hype available for sale. They are not tied to the DeepSeek brand officially, hence their capitalisation is speculative.
Such scam schemes like creating phishing emails are quite widespread with popular AI models - it’s important to recognize that cybercriminals will inevitably seek to exploit such tools for malicious purposes.
“What stands out in the case of DeepSeek is its open-source nature. While open-source frameworks foster transparency, collaboration, and innovation, they also introduce significant security and ethical risks.
“When you’re using an open-source tool, you can’t always be sure how your data is being handled, especially if someone else has deployed it,” comments Leonid Bezvershenko, Security Researcher, Kaspersky GReAT.
“Exploitation of open-source software was a major trend in the threat landscape last year, with cybercriminals running complex campaigns to embed malware. In 2024, our open-source scanner detected over 12,000 malicious packages in open repositories.
“Without centralised oversight, threat actors can start creating compromised versions of software or introduce backdoors under the guise of tools for using the API of DeepSeek, presenting serious risks to users and organisations.”
