In recent discussions about AI technologies, DeepSeek AI has emerged as a potential alternative to ChatGPT. However, before organizations consider integrating this model, it’s crucial to understand the associated risks. Below, we summarize key findings and concerns about DeepSeek AI, along with an alternative recommendation for those prioritizing data protection and transparency.

Security Risks Highlighted by Kela

The cybersecurity firm Kela, known for its expertise in monitoring cybercrime, has analysed DeepSeek AI and uncovered significant security vulnerabilities. Their findings, detailed in a recent blog post, reveal:

  • Potential for malicious use: DeepSeek AI can be exploited to generate harmful content, including malware, violating the principles of the EU AI Act.

Data Privacy Concerns

DeepSeek AI’s privacy practices raise additional red flags, particularly for organizations operating within the EU. The model is developed by a Chinese company, making it subject to Chinese laws. This creates the possibility that user data could be accessed by Chinese authorities. Moreover, the company’s privacy policy (accessible here)  includes several troubling points:

  • Data storage in China: User data is stored on servers located in China, which poses compliance challenges for EU organizations.
  • Lack of clarity on security measures: The policy does not provide detailed information on how data is protected during transfers.
  • Broad data-sharing permissions: User inputs can be used for advertising and analysis purposes, without sufficient transparency or the option to opt out.

Why Consider Mistral AI?

For organizations prioritizing privacy and regulatory compliance, Mistral AI stands out as a reliable alternative (Privacy Policy available here). As a European-based model, Mistral AI emphasizes:

  • Data protection: Ensuring user data remains secure and processed within strict European regulations.
  • Transparency: Offering clear policies and controls over data processing, giving organizations confidence in their AI usage.

Conclusion

While DeepSeek AI may appear to be a promising solution, the outlined risks and privacy concerns warrant careful consideration. For organizations seeking a secure and privacy-friendly AI model, Mistral AI offers a compelling alternative. If your organization requires further guidance on this topic, we are available to provide detailed consultations.

| | | , , , , | 5 Kommentare