The meteoric rise of DeepSeek, a Chinese AI startup, has not only disrupted the AI sector but also attracted the attention of cybercriminals.
Following the release of its open-source model, DeepSeek-R1, on January 20, 2025, the platform experienced exponential growth, reaching millions of users globally.
However, this popularity has been exploited by threat actors who have launched sophisticated phishing campaigns and malware attacks targeting unsuspecting users.
Cybersecurity researchers have identified numerous phishing websites mimicking DeepSeek’s official site.
These fraudulent platforms employ “site proxy” techniques to replicate the legitimate interface while embedding malicious links.
Instead of directing users to download the authentic app, these fake sites deliver malware-laden files.
One such phishing domain, deepsekk[.]sbs, was flagged as critically risky by cybersecurity tools, with antivirus programs detecting its payload as a malicious application targeting financial software.
Malware Impersonation on Developer Platforms
Threat actors have also leveraged Python Package Index (PyPI) to distribute malicious packages under names like “deepseeek” and “deepseekai.”
These packages impersonate developer tools for integrating DeepSeek into projects but are designed to steal sensitive data such as API keys and database credentials.
The stolen information is exfiltrated to command-and-control servers, enabling attackers to compromise cloud services and infrastructure resources.
In another campaign, cybercriminals used fake CAPTCHA verification schemes under domains like deepseekcaptcha[.]top to trick users into executing malicious PowerShell commands.
These attacks often result in the installation of infostealers like Vidar Stealer and Lumma Stealer, further compromising user data.
Security Concerns with DeepSeek’s AI Model
DeepSeek’s flagship model, DeepSeek-R1, has also faced scrutiny for its vulnerabilities.
Security tests revealed that the model is highly susceptible to adversarial prompts and jailbreaking techniques.
Researchers found that it could generate harmful content, including malware scripts and phishing templates, with a failure rate exceeding 90% in blocking such prompts.
This raises concerns about its safety mechanisms and potential misuse by malicious actors.
Additionally, DeepSeek’s mobile app has been criticized for weak encryption protocols and unencrypted data transmissions.
Security assessments revealed that sensitive user data is transmitted over insecure channels and stored on servers in China, raising privacy and regulatory compliance concerns.
The exploitation of DeepSeek’s popularity underscores the risks associated with rapidly scaling technologies without robust security measures.
Users are advised to verify domains before downloading applications or interacting with online platforms claiming affiliation with DeepSeek.
Developers should exercise caution when integrating third-party tools by validating package authenticity through trusted sources.
Organizations can mitigate these risks by employing threat intelligence tools like Criminal IP Domain Search to identify high-risk domains and pre-blocking malicious links using browser extensions.
Furthermore, enterprises should consider restricting the use of AI platforms like DeepSeek in critical workflows until security vulnerabilities are adequately addressed.
The case of DeepSeek highlights the urgent need for enhanced cybersecurity measures in the AI ecosystem as threat actors continue to exploit emerging technologies for malicious purposes.
Investigate Real-World Malicious Links & Phishing Attacks With Threat Intelligence Lookup - Try for Free
