Listen to this Post
2025-02-21
Chinese startup DeepSeek AI has been generating significant buzz this week, thanks to its powerful open-source language models that rival AI systems like Anthropic’s Claude and OpenAI’s GPT series. However, the release has sparked concerns regarding censorship, data privacy, and the influence of the Chinese government on its training. In response, AI platforms like Perplexity and You.com have found innovative ways to give users access to uncensored versions of DeepSeek’s models while addressing security risks. Here’s how you can try DeepSeek R1 without compromising on freedom or privacy.
Summary:
DeepSeek AI, a Chinese-based AI company, recently released its language model, R1, which has been compared to other advanced models like OpenAI’s GPT. However, concerns about censorship and data security have arisen because of DeepSeek’s ties to Chinese servers, where government-enforced censorship practices may influence the model’s responses.
To counteract these issues, Perplexity has launched an uncensored version of DeepSeek’s R1 model, called R1 1776, which is open-sourced and designed to provide unbiased, factual information. This model is available through Perplexity’s platform, HuggingFace, and its Sonar API, while the original censored version is still accessible on major platforms like Azure and AWS.
Perplexity offers users access to R1 on a free plan, with the option to upgrade to a Pro version for $20 per month to access more queries. Additionally, the company assures that data used through its platform remains within Western data centers, ensuring privacy.
You.com, another platform offering access to DeepSeek models, hosts R1 and V3 models on its Pro tier, which costs $15 per month. Users on You.com can choose to use the models with various trust layers, adjusting settings for web source access and custom agent configurations. Despite these safeguards, censorship remains a potential issue, particularly for politically sensitive topics.
What Undercode Says:
DeepSeek’s entrance into the AI market with its R1 model is notable for its advanced capabilities, but the controversies surrounding its censorship and data security pose significant concerns. The model’s connections to Chinese servers raise questions about the potential risks involved in using AI tools that are subject to government-imposed content restrictions. Despite these challenges, the move by platforms like Perplexity and You.com to offer uncensored and privacy-conscious versions of R1 showcases the flexibility of open-source AI and highlights the potential for mitigating some of the risks associated with DeepSeek’s models.
From a privacy standpoint, Perplexity’s assurance that all data stays within Western data centers is a positive step toward reducing the risk of data exposure to Chinese authorities. This commitment is crucial, especially for users in regions with stringent data privacy regulations such as the EU and the US. However, while the censorship issue seems to have been addressed to some extent, users should remain cautious when it comes to politically sensitive topics, as R1’s training model is inherently tied to the Chinese government’s censorship practices. For example, when asked about the Tiananmen Square protests, R1 refused to respond, even on a supposedly uncensored platform, suggesting that censorship may not be entirely avoidable, depending on how the model is deployed.
On the other hand, You.com’s approach to allow users to adjust the AI’s access to public web sources gives users more control over the information that the model can access. This customization not only helps mitigate censorship concerns but also enables users to explore the model’s behavior in a way that aligns more with their specific needs. By removing public web access, users can limit biases that could stem from reliance on web data, though this approach might also limit the model’s accuracy in certain contexts.
Another critical aspect of the DeepSeek R1 model is its performance in various areas, such as coding. Tests have shown that the R1 model performs well when it comes to technical tasks, and it could be an excellent resource for developers who need reliable code-generation assistance. However, users must remain vigilant about what information is being censored, as the AI’s design may still carry underlying biases, despite the decensoring process.
While some may find the political biases in DeepSeek’s model a serious drawback, it’s worth noting that these models are still evolving. As the AI landscape matures and more tools become available for customization and control, platforms like Perplexity and You.com are leading the charge in creating environments where users can test, modify, and control the AI’s responses. This shift indicates a future where users might have more say in the ethical deployment of AI technologies, especially regarding issues of censorship and privacy.
For those seeking an AI tool that offers the flexibility of open-source models but with safeguards against political bias and data privacy violations, DeepSeek R1 on platforms like Perplexity and You.com offers a viable solution. By taking advantage of these options, users can leverage powerful language models without the concerns of unwanted censorship or exposure to governmental surveillance, marking an important milestone for the future of open-source AI.
References:
Reported By: https://www.zdnet.com/article/perplexity-lets-you-try-deepseek-r1-without-the-censorship-or-security-risk/
Extra Source Hub:
https://www.digitaltrends.com
Wikipedia: https://www.wikipedia.org
Undercode AI
Image Source:
OpenAI: https://craiyon.com
Undercode AI DI v2
