Is Fear Holding You Back? The Truth About DeepSeek R1 and AI Innovation

Due to the timeliness of this information, I am publishing an extra article this week.

Just a few days prior to the release of DeepSeek R1, many people, including a number of my colleagues, were touting that Project Stargate ?had cemented America’s position as the world’s AI superpower, perhaps for the next half-century.

The AI War Just Got Interesting

Then a small Chinese startup ?DeepSeek made headlines with its new AI model, R1, which rivals leading U.S. models including the latest reasoning models from OpenAI, but was developed at a fraction of the cost and using older (and cheaper) Nvidia chips. Talk about disrupting an industry!

I keep hearing people say they won’t use DeepSeek because it’s Chinese. And I get the concern—there are real issues when it comes to tech and foreign influence. That’s why TikTok has been all over the news. It’s not something to ignore. In fact, here are a few key sentences from DeepSeek’s Terms and Conditions and Privacy pages:

• DeepSeek products and services are jointly owned and operated by Hangzhou DeepSeek Artificial Intelligence Co., Ltd., Beijing
• these Terms shall be governed by the laws of the People’s Republic of China
• When you use our Services, we may collect your text or audio input, prompt, uploaded files, feedback, chat history, or other content that you provide to our model and Services.
• We automatically collect certain information from you when you use the Services, including internet or other network activity information such as your IP address, unique device identifiers, and cookies.
• if you choose to sign-up or log-in to the Service using a third-party service such as Apple or Google, or link your account to a third-party service, we may collect information from the service
• We retain information for as long as necessary to provide our Services and for the other purposes set out in this Privacy Policy.
• The personal information we collect from you may be stored on a server located outside of the country where you live. We store the information we collect in secure servers located in the People’s Republic of China.

Sounds scary, right? Then again, have you read the privacy policy of the other software tools you are using? But here’s what got me thinking…

I’m using an Android phone. My laptop is an HP. And while neither one is made in the U.S., I still rely on them every day. If we avoided every product with ties to China, most of our electronics—and a whole lot of other things including the clothes on our back—would be off the table.

Now, here’s the part that matters: DeepSeek is open-source. That means there’s no hidden control, no mystery about how it works. For example, HuggingFace.co has already announced the launch of their Open-R1 project, a fully open reproduction of DeepSeek R1. If security is the issue, the solution is transparency—and DeepSeek provides that. You can run it on your own servers, just like any other open AI model.

The bigger question is: are we letting fear hold us back from using powerful tools? As the AI Architect for Business Success, I realized that I must do my own testing and not just rely on what others were saying. It would be irresponsible of me to not conduct my own evaluation of this tool.

Here is what I have discovered:

• When I tried to create an account with my email address, I saw two notices. The first said “Due to large-scale malicious attacks on DeepSeek’s services, registration may be busy. Please wait and try again.” Okay, I get that the scammers of the world want to hack into anything they can. But his does raise security concerns.
• What was more concerning was the message: “Error sending code. Your email domain is currently not supported for registration.” I was forced to use a gmail email address. When I did do that, I got a message: “Signup failed.” This happened multiple times. (perhaps due to message #1).
• Yes, it is censored to avoid making any comments about China and its government. Personally, I don’t create content about China so that isn’t an issue for me. But if they are censoring that information, what else might it be hiding?
• The release of DeepSeek-R1 is an amazing boon for the AI community, not a devasting blow as some make it out to be. Sure, Nvidia stock has taken a hit and people are questioning their large investments into companies like OpenAI and Anthropic, but these organizations can build upon R1 for their own use. The kicker is DeepSeek didn’t release everything—although the model weights are open, the datasets and code used to train the model are not. That’s what HuggingFace is attempting to provide.

The results:

Using the exact same prompt: “Summarize the most significant global events from the past week, highlighting key developments in politics, economics, technology, and international relations. Provide an analysis of their potential long-term impact and include credible sources where possible. Additionally, assess how different media perspectives may influence public perception of these events.”

The exact responses to this prompt are posted here. Below is a summary if my findings.

Chat GPT – provided a concise answer to my prompt, including real-time information and reference sources.

Google Gemini – more wordy, but generally good information with sources. Decent image generation.

MS CoPilot – the most wordy, but detailed information with sources. Image generation is decent but not perfect (hands with 3 fingers and misspelled words).

Perplexity – Good layout and information, including sources, but missed the story about DeepSeek. While known for being an AI search engine, I have often found that Perplexity misses information I’d expect in a search. Unable to create images.

Anthropic Claude – No real time access to the internet, so answer was insufficient. Image generation is weak. Claude is known for have one of the better writing styles however and is often thought of (at least prior to DeepSeek) as the main competitor to OpenAI.

DeepSeek – The layout looks eerily similar to ChatGPT. The output was generally good. It is unable to create images but gave a prompt and instructions for how to do this using tools like Adobe Illustrator or Canva.

Since this is an article about DeepSeek, I did several other tests just using it to see how well it performed. Overall, it does a good job of writing. It is not yet capable of remembering things, so training it to write like you based on previous work would require uploading all of your documents each time. However, the output of the writing it provided sounded compelling and convincing.

I tested it with some creative fictional writing as well. It compared favorably to ChatGPT. I wouldn’t say one is better than the other – they were both pretty good. With a little human oversight and the back-and-forth prompting to tweak the output as is often required when chatting with an AI.

As for reasoning, I provided several reasoning types of questions to both DeepSeek and ChatGPT o1. Both correctly identified the proper answer. The fact that DeepSeek has been able to create a reasoning model as fast and as accurate as OpenAI is amazing, especially since you aren’t paying for it.

My thoughts on this are that DeepSeek is not as versatile of a tool as ChatGPT. I’m not giving up my subscription to ChatGPT anytime soon. And we all know that whenever something is “free” like DeepSeek (or Google) is, that you, the consumer, are the product. Your information will be used and sold (including to China). Clearly, you don’t want to use any sensitive information as a result and you need to be okay with that.

At the same time, AI users are the real winner here. Sure, Nvidia and other tech giants may be scratching their heads wondering how they could possibly be the ones whose business model got disrupted, but for consumers, driving down the price of AI is a positive thing.

What do you think? Are you using DeepSeek, or have concerns kept you from trying it?