Microsoft & Perplexity Give DeepSeek Their Stamp of Approval

Microsoft (via Azure AI Foundry and GitHub) and Perplexity (for Pro subscribers).,Concerns remain over data privacy, with Wiz discovering a vulnerability exposing over a million records.,Censorship fears plague DeepSeek’s official version, but Perplexity claims to have bypassed those issues.,Microsoft asserts that its evaluation process ensures a “secure, compliant” environment for enterprise users, while OpenAI accuses DeepSeek of training on its models.,The model is cheap, powerful, and quickly gaining popularity—raising questions about broader implications for AI innovation and geopolitics.

DeepSeek R1 in Microsoft & Perplexity Integrations

Hello, dear readers of AIinASIA! Buckle up because today's story is a whirlwind of controversy, tech breakthroughs, and a dash of political intrigue. If you've ever wondered about the future of Chinese-developed AI models on the global stage, you’re in for a treat. In today’s article, we’re talking about a certain trailblazing model called DeepSeek R1—and how Microsoft and Perplexity are making waves by integrating it into their platforms, despite the uproar surrounding data privacy and censorship concerns.

Microsoft & Perplexity Give DeepSeek Their Stamp of Approval

Let’s dive in: DeepSeek R1—fresh off the press just 10 days ago—has already found a home with some American tech powerhouses. Both Microsoft and Perplexity have integrated DeepSeek’s model into their offerings. On Microsoft’s side, you can find R1 on Azure AI Foundry (for subscribers) and on GitHub (free to access). Perplexity, meanwhile, is providing DeepSeek R1 to its Pro subscribers (currently going for $20/month), putting it alongside other AI heavy-hitters like OpenAI’s GPT-o1 and Anthropic’s Claude-3.5.

But hang on—this isn’t just a simple case of “Chinese model meets US platforms.” DeepSeek had plenty of privacy and censorship red flags raised against it. So, how did these American companies feel confident enough to go ahead with it?

Well, according to Microsoft, the company conducted “rigorous red teaming and safety evaluations.” Its official stance is that by accessing DeepSeek R1 through Microsoft services, users get a “secure, compliant, and responsible environment for enterprises.” In other words, Microsoft is telling the world, “No worries—our version is safer!”

Perplexity echoes that sentiment, boldly claiming its version of DeepSeek isn’t censored by the Chinese government. In fact, Perplexity CEO Aravind Srinivas even took a jab at the default DeepSeek chatbot, pointing out how it refused to discuss topics like the Tiananmen Square Massacre or the treatment of Uyghurs. And if you tried talking to it about Taiwan—the responses suspiciously sounded like they came straight from Chinese Communist Party (CCP) talking points. Perplexity says it has completely dodged those issues by tweaking the model to keep it “off-script.”

Data Privacy Snags & Vulnerabilities

Just when we thought things were cruising along nicely, a security research firm called Wiz dropped a bombshell. It discovered a DeepSeek data vulnerability exposing over a million records—ranging from sensitive data to chat logs—publicly on the web. Talk about a privacy nightmare!

Wiz disclosed the problem to DeepSeek, which then patched the flaw. But another data-collection concern remains: If you chat on the official chat.deepseek.com website, there’s every chance the Chinese government could lay its hands on your data. For many, that’s simply a risk not worth taking.

Thankfully, not everyone is using DeepSeek in a way that might send your info across the ocean. According to Endor Labs, an open-source security company, hosting the model yourself (such as grabbing it on Hugging Face) doesn’t pose the same kind of data-sharing risk. Microsoft is even hinting that a “distilled” version of DeepSeek could soon be coming to Copilot+ PCs for local usage—promising better speed, privacy, and efficiency.

OpenAI’s Not Happy

Here’s the spicy part: OpenAI has accused DeepSeek of secretly training on its own models. Although OpenAI hasn’t gone into detail, former US “AI Czar” David Sacks told Fox News there’s “substantial evidence” to back up that claim. Critics, however, have pointed fingers at OpenAI’s own data-gathering practices, saying it, too, has used “stolen IP.” What’s that saying about people in glass houses?

Whatever the truth, you can see why OpenAI might not be thrilled about DeepSeek’s rising popularity. If a brand-new model from a competitor can match or outperform established giants, it puts pressure on the entire ecosystem to innovate faster. Indeed, the AI arms race is a hot topic, forcing rapid advancements.

So, What’s Next?

DeepSeek R1 doesn’t look like it’s going away anytime soon—especially now that big names like Microsoft and Perplexity have legitimised it. We can likely expect this model to spread even further across the AI landscape, given its blend of low cost, high performance, and the possibility of skipping that pesky censorship if you use the right channel. This aligns with broader trends in APAC AI in 2026.

Yes, there are still caveats—data vulnerabilities and political entanglements are serious considerations. But for many AI enthusiasts and developers, the allure of accessing a powerful reasoning model at a fraction of typical costs is too hard to resist. Who wouldn’t want something that can reportedly “think out loud like an intelligent person” and process “hundreds of sources” in one go? This also raises questions about AI's trust deficit in Southeast Asia.

What Do YOU Think?

Are we too quick to embrace powerful AI models from abroad—and in doing so, are we opening the door to government-level surveillance and censorship creeping into the global tech ecosystem? This is a critical discussion, especially as AI's wave shifts to the Global South.

As always, keep your eyes peeled right here on AIinASIA for more updates! We’ll be diving deeper into how these AI skirmishes play out on the global stage—because if there’s one thing we love around here, it’s a good AI battle royale.

Let’s Talk AI!

How are you preparing for the AI-driven future? What questions are you training yourself to ask? Drop your thoughts in the comments, share this with your network, and Subscribe to our newsletter for more deep dives into AI’s impact on work, life, and everything in between.

YOUR TAKE

We cover the story. You tell us what it means on the ground.