Perplexity AI, a San Francisco-based AI search engine, has announced the integration of the Chinese DeepSeek R1 reasoning model into its Pro-level offerings. This addition allows Pro users to leverage the model for advanced research capabilities, including uncensored responses and local data security.
DeepSeek R1 is available as a self-hosted, open-source model hosted in secure data centers located in the USA and Europe. Perplexity AI’s cofounder and CEO, Aravind Srinivas, clarified misconceptions surrounding the launch, including concerns about potential data sharing with the Chinese government and censorship. Srinivas emphasized that user data remains entirely within local environments, with no transmission to external parties.
In response to censorship concerns, Srinivas shared examples of uncensored responses from DeepSeek R1, demonstrating that the model can be customized to operate without restrictions when self-hosted. This reassurance addresses worries about the model’s default censorship, a characteristic attributed to its original training data.
While third-party testing has confirmed that the base DeepSeek R1 model includes content limitations, Perplexity AI’s implementation ensures users can fully control and adapt the tool to their needs.
This latest addition reinforces Perplexity AI’s commitment to offering diverse, customizable solutions for its users, positioning the platform as a leader in the AI-powered search space.