Perplexity Lets you Try DeepSeek R1 without the Security Risk, however it’s Still Censored

Chinese start-up DeepSeek AI and its open-source language models took over the news cycle today. Besides being comparable to designs like Anthropic’s Claude and OpenAI’s o1, the designs have actually raised a number of issues about data personal privacy, security, and Chinese-government-enforced censorship within their training.

AI search platform Perplexity and AI assistant You.com have actually discovered a way around that, albeit with some constraints.

Also: I evaluated DeepSeek’s R1 and V3 coding abilities – and we’re not all doomed (yet)

On Monday, Perplexity published on X that it now hosts DeepSeek R1. The totally free strategy provides users 3 Pro-level questions per day, which you could use with R1, but you’ll require the $20 each month Pro strategy to gain access to it more than that.

DeepSeek R1 is now offered on Perplexity to support deep web research. There’s a new Pro Search thinking mode selector, together with OpenAI o1, with transparent chain of thought into design’s reasoning. We’re increasing the number of day-to-day usages for both complimentary and paid as add more … pic.twitter.com/KIJWpPPJVN

In another post, the company verified that it hosts DeepSeek “in US/EU information centers – your data never leaves Western servers,” assuring users that their information would be safe if using the open-source designs on Perplexity.

“None of your data goes to China,” Perplexity CEO Aravind Srinivas restated in a LinkedIn post.

Also: Apple researchers expose the secret sauce behind DeepSeek AI

DeepSeek’s AI assistant, powered by both its V3 and R1 models, is accessible via browser or app– but those need communication with the company’s China-based servers, which produces a security danger. Users who download R1 and run it on their gadgets will prevent that problem, however still face censorship of particular topics determined by the Chinese federal government, as it’s integrated in by default.

As part of offering R1, Perplexity declared it eliminated a minimum of some of the censorship built into the design. Srinivas published a screenshot on X of question results that acknowledge the president of Taiwan.

However, when I asked R1 about Tiananmen Square utilizing Perplexity, the model declined to address.

When I asked R1 if it is trained not to respond to specific questions identified by the Chinese federal government, it reacted that it’s developed to “focus on accurate information” and “avoid political commentary,” which its training “stresses neutrality in worldwide affairs” and “cultural level of sensitivity.”

“We have actually eliminated the censorship weights on the model, so it shouldn’t behave this way,” stated a Perplexity agent reacting to ZDNET’s ask for remark, including that they were checking out the problem.

Also: What to understand about DeepSeek AI, from cost claims to data privacy

You.com uses both V3 and R1, likewise just through its Pro tier, which is $15 each month (marked down from the normal $20) and without any totally free inquiries. In addition to access to all the models You.com uses, the Pro plan includes file uploads of approximately 25MB per query, a 64k optimum context window, and access to research and custom representatives.

Bryan McCann, You.com cofounder and CTO, described in an email to ZDNET that users can access R1 and V3 by means of the platform in 3 methods, all of which use “an unmodified, open source version of the DeepSeek models hosted totally within the United States to guarantee user personal privacy.”

“The very first, default method is to utilize these models within the context of our proprietary trust layer. This provides the models access to public web sources, a predisposition towards mentioning those sources, and a disposition to respect those sources while producing responses,” McCann continued. “The second method is for users to turn off access to public web sources within their source controls or by utilizing the models as part of Custom Agents. This option enables users to explore the models’ distinct abilities and habits when not grounded in the general public web. The third way is for users to evaluate the limitations of these designs as part of a Custom-made Agent by including their own guidelines, files, and sources.”

Also: The best open-source AI models: All your free-to-use choices discussed

McCann noted that You.com compared DeepSeek designs’ reactions based upon whether it had access to web sources. “We observed that the models’ actions differed on numerous political topics, sometimes declining to respond to on particular problems when public web sources were not consisted of,” he describes.