DeepSeek AI: Data Privacy and Security Concerns
Introduction
Chinese startup DeepSeek AI and its open-source language models have recently taken the news cycle by storm. The models have raised several concerns about data privacy, security, and Chinese-government-enforced censorship within their training.
Perplexity and You.com Offer Alternative Options
AI search platform Perplexity and AI assistant You.com have found a way to overcome these concerns, albeit with some limitations. Perplexity now hosts DeepSeek R1, a free plan offering three Pro-level queries per day, with a $20 per month Pro plan for more access. You.com, on the other hand, offers both V3 and R1 models only through its Pro tier, which costs $15 per month and includes additional features like file uploads and custom agents.
Data Security and Censorship Concerns
Perplexity CEO Aravind Srinivas assured users that their data would be safe, stating, "None of your data goes to China." However, the DeepSeek AI assistant, powered by V3 and R1 models, requires communication with China-based servers, creating a security risk. Users who download R1 and run it locally on their devices will avoid this issue, but may still encounter censorship of certain topics determined by the Chinese government.
Perplexity’s Solution
Perplexity removed at least some of the censorship built into the model, as shown in Srinivas’ LinkedIn post. However, when asked about Tiananmen Square, the model refused to answer. When questioned about being trained not to answer certain questions determined by the Chinese government, R1 responded that it is designed to "focus on factual information" and "avoid political commentary," and that its training "emphasizes neutrality in global affairs" and "cultural sensitivity."
You.com’s Solution
You.com cofounder and CTO Bryan McCann explained that users can access R1 and V3 models via the platform in three ways, all of which use an unmodified, open-source version of the DeepSeek models hosted entirely within the United States to ensure user privacy. The models can be used with or without public web sources, allowing users to explore their unique capabilities and behavior.
Conclusion
The concerns surrounding DeepSeek AI’s data privacy and security have raised important questions about the use of these models. Perplexity and You.com’s solutions offer a way to access these models while maintaining some level of control over data and censorship. However, more research is needed to fully understand the implications of these models and their potential biases.
FAQs
Q: Is my data safe when using Perplexity’s DeepSeek R1?
A: Yes, according to Perplexity CEO Aravind Srinivas, your data will be safe, as it is processed entirely within Western servers.
Q: What are the limitations of using DeepSeek AI?
A: The models require communication with China-based servers, creating a security risk, and may encounter censorship of certain topics determined by the Chinese government.
Q: How can I access DeepSeek AI models?
A: Perplexity and You.com offer Pro plans that include access to DeepSeek R1 and V3 models. Perplexity’s free plan offers three Pro-level queries per day, while You.com’s Pro plan costs $15 per month.
Q: Can I use DeepSeek AI models without including public web sources?
A: Yes, You.com allows users to turn off access to public web sources within their source controls or use the models as part of Custom Agents, giving them control over the models’ behavior.
