Is Deepseek Spyware

Deepseek have created an earthquake throughout Nasdaq and the US Government the last couple of days. Not only did they outperform OpenAI, but they also released everything as open source. It's therefore frustrating for me to read how some people are claiming Deepseek is spying on them. So let's clarify the facts here.

API versus Model

Deepseek is a Chinese company. According to Chinese laws, they have to give all data they collect to the Chinese Communist Party (CCP). This implies that if you're using their chat or their API, the CCP will read everything you write. However, Deepseek also released their models as open source. This allows companies such as AINIRO to download their models, host it on our own hardware, and deliver solutions to our clients with less spying than if we were to use OpenAI.

This is because of that at this point it's no longer going through Deepseek's API, but through our API - And the model itself is not even in theory capable of transmitting data to China, unless we explicitly implement it - Which we of course never will do unless ordered to do so by a Cypriot court of law, which is highly unlikely.

Since we are a Cypriot company based in EU, we are not subjects to the US Patriot Act. This implies that with an LLM hosted by AINIRO, you would actually experience less government interference, because of EU's slightly more relaxed laws when it comes to coerced surveillance by proxy.

Propaganda concerns

Deepseek's models refuse to answer questions related to Tiananmen Square. This makes it biased. However, there's no more bias in Deepseek than in OpenAI, it's just different bias. If you don't believe me, try asking ChatGPT about Korona or Covid.

Is bias a problem? Yes, but we can deliver AI solutions based upon Deepseek that answers questions about Tiananmen Square. However, we cannot deliver AI solutions based upon OpenAI that answers questions about Korona or Covid accurately. This is because we can download Deepseek, fine tune it with new data, and change its answers accordingly. This is not an option with OpenAI for us. So the "propaganda problems" are arguably larger for us with OpenAI than they are with Deepseek.

OpenAI answering questions about Covid

And yes, I know OpenAI allows for fine tuning their models, but then the resulting model is still hosted by OpenAI - Implying their "censorshop filters" would still apply, making it literally impossible to deliver an AI solution answering questions about Korona accurately. While delivering an AI solution based upon Deepseek that answers questions about Tiananmen Square correctly would be quite easy for us.

Potential Motivations

It's always dangerous to guess other peoples' motivations. For all I know it might be as simple as that Xi Jinping was drunk at a party and wanted to impress his girlfriend by disrupting US economy to such an extent that half of Nasdaq top 100 went bankrupt - Which is a legitimate potential consequence of this may I add.

However, I find it more likely that he simply wants to apply some "disruptive destruction". US tech companies have increased their evaluations 10x since AI became a thing. By damaging these companies where it hurts, the CCP can win "an ideological quarrel", that results in us (the people in the west) becoming more friendly towards them in general, resulting in that the CCP wins "propaganda points".

It could also very well simply be revenge because of the TikTok lockout, where the objective is to "coerce" the DHS to allow for Chinese companies to operate in the US.

US and China is in an ideological battle, and in this battle they are fighting for the hearts and minds of men, on both sides. My piece of advice is to just let them continue fighting, because the only way they can possibly score "propaganda points" is by giving more power to us. And ...

As long as they're fighting each other, they won't have time to fight us! 😎 🥳

Wrapping up

I am not a communist, and I dislike government interference in general. This is true for both the US government and the CCP. I think we'd all be better of without any of these in our lives. However, if you're one of those running around screaming that the CCP will use Deepseek to surveillance us, you're completely missing the point.

In fact, I'd be very surprised if OpenAI did not build GPT5 on top of lessons learned from Deepseek. So whether or not Deepseek stole from OpenAI is also irrelevant. It could be argued that most engineers working at Google and OpenAI are Indians, and that the US therefore stole OpenAI from India. The important question is how much OpenAI will (and can) steal from Deepseek to deliver their own next model. And with open source "stealing" is perfectly legal - It's the way the game works.

However, the long term effects of this is dramatic, because the promise is that AGI and ASI will not be exclusively provided to the few rich, but rather will become something we all can have a share in. Deepseek allows us to run our own models locally, based upon Deepseek, that basically outperforms the currently 500 billion dollar investment that Donald Trumpt promised to AI research in the US. At the end of the day, it's really as simple as the following ...

The CCP just gave us all 500 billion dollars each. Do you want to accept them, or do you want to refuse it "because of their ideological motives" ...?

Sometimes what's good for you, is not necessarily what's good for your government and our western companies. And this is definitely one of those things. As for me, I'll take my 500 billion dollars, say thank you, and build my own LLMs based upon Deepseek! If you want to run around screaming about Chinese surveillance, than I have no problem with that. In fact, the more you do that, the more of your 500 billion dollars I can snatch out of your hands because of your prejudice, ignorance, and paranoia. As for me however - I've got only one thing left to say, which is as follows ...

Thank you Deepseek! 🥰

Have a Custom AI Solution

At AINIRO we specialise in delivering custom AI solutions and AI chatbots with AI agent features. If you want to talk to us about how we can help you implement your next custom AI solution, you can reach out to us below.