DeepSeek Allegedly STOLE OpenAI’s Tech... But New AI JUST BEAT Them Both!

Microsoft and OpenAI are now investigating DeepSeek over claims that they stole OpenAI's tech. And this comes right after the Chinese Startup dropped their R1 model, a powerful AI that supposedly cost just a fraction of what OpenAI spent on GPT4, yet performs just as well, if not better, than OpenAI's 01. And just when you think it couldn't get any crazier, Alibaba jumps in, claiming their brand new Quinn 2.5 model beats both Deepseek and OpenAI on benchmarks. Yeah, the AI wars just got real. So let's break it all down and see how this wild saga is playing out. Alright, so first off, why is Microsoft, one of OpenAI's biggest investors, so concerned? Well, according to reports from Bloomberg, Microsoft's security researchers saw suspicious activity tied to a group they believe is is connected to Deep Seek. It all happened in the fall of last year when these individuals were allegedly pulling out huge amounts of data from OpenAI's API. By the way, that's the same API used by developers who pay OpenAI to power their own apps and services. So it's a major revenue channel. As soon as Microsoft picked up on those red flags, they tipped off OpenAI and the investigation started. Now the big question is, did Deep Seek actually see steal anything valuable from OpenAI's models? David Sacks, the White House's so called AI and crypto czar, told Fox News that there's substantial evidence that Deepseek basically distilled OpenAI's knowledge into its own system. Distillation, if you're not super familiar with that AI term, is basically a technique where one model learns from the outputs of another, effectively compressing or copying some capabilities. And so if that's really what happened, yeah, that's a big no no from an intellectual property standpoint, especially with tensions running high between the US and China over advanced tech. And get this, earlier in January, Deepseek's free AI assistant soared past ChatGPT to become the top free app on Apple's App Store in the us and it's also number one in dozens of other countries. That kind of success obviously raises eyebrows. Like how did this Chinese startup, which apparently came out of nowhere, manage to beat all these American AI heavyweights on the download charts. It's a story big enough to trigger an actual sell off of tech stocks on Monday. With investors scrambling to figure out if the market's illusions about who dominates AI might be shattered, deepseek's core claim is that they've built a large language model referred to as R1 that has reasoning capabilities competitive with the best OpenAI can offer, despite spending nowhere near the same billions of dollars. In fact, the company put out a research paper saying it only spent about $5.6 million to train the underlying foundation model, called V3, using 2000 Nvidia H800 GPUs. Just for comparison, Sam Altman over at OpenAI hinted that GPT4 might have cost well over $100 million and used maybe 25,000 H100 GPUs. So how on earth does Deep Seek get to that same level with such fewer resources? Not everyone is buying it. For instance, Palmer Luckey, founder of Oculus VR, called DeepSeek's numbers bogus, going so far as to brand them Chinese propaganda. He even suggested it might be a psyop designed to disrupt US Tech investment or manipulate certain short positions in the stock market. Elon Musk, who's super close to President Donald Trump, also said he's skeptical, basically chiming in on social media with the firm, obviously when someone suggested Deepseek had hidden access to advanced hardware. Meanwhile, Alexander Wang from Scale AI believes that Deepseek probably got its hands on way more advanced hardware, like 50,000 H100 GPUs, in secret. He didn't provide solid evidence, but the rumor is strong enough to keep fueling speculation that DeepSeek's shoestring budget story might be leaving out some details. At the end of the day, you have to wonder, are these companies genuinely concerned about stolen tech, or are they just bitter that a smaller player outperformed them without needing billions in funding? Deepseek didn't ask investors for unlimited cash, didn't follow the same playbook as OpenAI or Google, yet somehow managed to create an AI model that competes at the highest level. And when you look at who's making these accusations, it's mostly companies whose own AI projects just got outshined, so it's not hard to see why they'd be quick to throw shade, whether it's about IP theft or questioning how Deepseek pulled it off so efficiently. And speaking of uncertainty, job security is more fragile than ever, making it crucial to stay ahead of with skills that can future proof your income. That's why we've teamed up with Growth School for this video to bring you something truly valuable that could help you thrive in this rapidly evolving landscape. 2024 has been a whirlwind, jobs popping up everywhere, but layoffs are just as common. It's a wild ride, and even if things feel secure now. You never know what's around the corner. That's why I think having multiple streams of income isn't just smart, it's essential. Here's where AI comes in with the right tools and skills, you could seriously start earning an extra $10,000 a month. Now, if you're wondering how to get started, Growth School has something really cool. They're offering a three hour hands on AI training where you'll learn to use over 25 powerful AI tools. Normally it's paid, but the first 1000 AI Revolution viewers can join for free using the link in the description. On top of that, you'll get $500 worth of bonus resources just for signing up. The training covers everything job hunting tips, salary negotiation, mastering Excel, even content creation. And it's not just for tech experts. Whether you're in finance, sales, marketing, HR, or even still studying, this can work for you. GrowthSchool has already helped millions of people level up and this could be your turn to stay ahead in an AI driven world. So if this sounds like your kind of thing, hit the link below to grab your free spot. Plus, don't miss joining Growth School's WhatsApp community. It's a great place to connect with others diving into AI too. But back to the AI showdown While the controversy rages on, real world benchmarks tell us a lot about how Deep Seek actually stacks up against the competition. Still, we do have some direct comparisons testing how well Deepseek's R1 handles tasks side by side with ChatGPT's latest O1 model and O1 Pro, and the results look basically competitive. One test measured each model's ability to do creative writing, math, logic, and so on. Turns out Deepseek R1 outperformed ChatGPT on some tasks, like identifying the 1 billionth prime number or writing certain creative paragraphs, while it lagged a bit on others, like certain complicated instructions or hidden code challenges. The takeaway is that R1 is in the same ballpark as OpenAI's best, which is pretty wild if you think about it, given how new Deepseek is to the global scene. As if that's not enough drama, we found out that part of Deepseek's success might be thanks to Huawei's hardware. Their R1 model was trained on Nvidia H100 or H800, but for inference, basically the in action stage when the model responds to user queries. It's apparently running on Huawei's Ascend 910C C AI chips. Now these chips aren't as powerful as Nvidia's top shelf ones, but you don't need as much raw computing for inference. So that means training can still rely on Nvidia, but once you've got the trained model, you can run it on cheaper or less regulated hardware. This detail also has big geopolitical ramifications, given that the US is trying to restrict China's access to advanced chips. After all these bombshells, the AI wars took another turn. Alibaba jumped on the bandwagon and announced Quen 2.5, an AI model that the company says can beat both GPT4O and DeepSeek V3 on various benchmarks. They also launched a model called Quen 2.5 VL, which they say handles complex text, image analysis, video analysis. Basically a super versatile AI. And guess what? As soon as Alibaba made this announcement, their stock went up nearly 3%. That's a stark contrast to how Nvidia's stocks reacted when Deep SEEK dropped its R1 model and basically wiped out $600 billion of Nvidia's market cap in a single day, the biggest single day loss in U.S. history. Talk about whiplash. Alibaba also claims it's training even bigger models like Quin 2.5 max on a whopping 20 trillion tokens, combining supervised fine tuning with something called reinforcement learning from human feedback. If you're thinking that sounds suspiciously similar to how OpenAI does it, you're not alone. But at the same time, all these companies are using these well known techniques Now, Alibaba's difference might be more about the scale and the ecosystem. They own one of China's largest cloud businesses, which pulled in more than $4.2 billion in revenue in the second quarter of 2025. So they definitely have the muscle to keep pushing this tech. To top things off, ByteDance, owner of TikTok, saw the perfect opening and decided to release an update to its own flagship AI model just days after Deepseek R1 came out. They're also claiming better performance than existing US systems. So yeah, it's basically an AI arms race in China right now. Everyone is competing to prove that they can develop powerful or at least very costafoovish AI without the monstrous budgets that OpenAI or Google have had. Meanwhile, we can't forget about the user side. Deepseek is at the top of the iOS app store in 51 countries. People have tested out the R1 chatbot and say it's pretty, pretty impressive, though it censors certain topics at the same time OpenAI's ChatGPT is still massive, especially among enterprise customers, and it's not like everyone's gonna jump ship overnight. Plus, ChatGPT has the advantage of Microsoft's huge infrastructure. There's also the bigger geopolitical angle. The US government has put restrictions on exporting advanced AI chips to China, but clearly even restricted chips like like the Nvidia H800 or Huawei's Ascend 910C still get the job done to some extent. That begs the question, is it really possible to hamper China's progress in AI by limiting top tier chips, or do they just find ways to make do with slightly inferior hardware but more efficient software? If the latter is true, that's obviously a headache for the US stance on trade and technology control. From a purely technical standpoint, one of the cooler comparisons we saw was the test where DeepSeek R1 and ChatGPT tried to figure out the billionth prime number. ChatGPT basically gave a rough theoretical estimate because that's how it's usually done, some prime number theorem plus a margin of error. But deep seq R1 cited actual references like Prime Grid stating the exact value 2020-2,801,763,489. That's a major wow moment for folks who assumed an American model would be the more Precise. Then again, R1 messed up some simpler instructions, like a puzzle where you have to encode letters in a hidden message by using the second letter of each sentence. So it's not perfect anyway. The biggest bombshell More recently is Alibaba's claim that that its brand new version of QEN 2.5 or Quen 2.5 Max outperforms GPT4O, DeepSeek V3 and Llama 3.405B across the board. That's a direct quote from the company. This statement comes right after all this mania around Deepseek's cheaper approach. It's definitely throwing down the gauntlet in the Chinese AI space, especially after ByteDance also hopped in. Some analysts think Alibaba timed its release for maximum attention. It was Lunar New Year in China, so people were at home with family and probably browsing the Internet or phone apps that guaranteed that Alibaba's announcement wouldn't get drowned out on a typical busy work day. So where does this leave us? Basically, the entire AI landscape is in flux. We have the big established players, Microsoft backed OpenAI, Google Meta, and we have challengers from China like Deepseek, ByteDance, and Alibaba the largest question mark is whether Deepseek actually stole intellectual property or used some sneaky method to replicate OpenAI's top level model. The investigation is ongoing and Microsoft and OpenAI haven't dropped any official bombs yet. But clearly Washington is aware and David Sachs is openly stating that Deep Sea Seek probably gleaned knowledge from GPT4's outputs. If that's proven, or if some settlement is reached, we can expect big headlines. Meanwhile, the AI arms race keeps charging ahead, and it's now not just about who can build the best model, but who can do it cheaply, who can roll it out faster, and who can handle the public's concerns around data security and censorship. The bottom line is this the AI sector is absolutely bustling with new players and controversies. Deepseek's alleged data breach is a testament to how valuable these large language models are. If it took OpenAI an estimated $100 million or more to train GPT4 and someone else can replicate it for under $6 million, that's a game changer. Assuming it's legit. But Alibaba's new claim, based on basically saying we just beat both OpenAI and DeepSeek, adds even more spice. Clearly the industry is bracing for more breakthroughs, more skepticism and a whole lotta competition. I'll be keeping a close eye on the Microsoft OpenAI Vs. Deepseek investigation because whatever happens next could reshape how these companies approach security and large scale licensing of their models. And definitely expect to hear more about Alibaba's Q QN models, especially since their stock is trending upward. We'll see if these claims hold up once the dust settles and people do more thorough tests. Alright guys, that's it for today's breakdown. It's insane how quickly the AI wars can shift. One minute you think only an American giant can lead the pack, then a Chinese startup or a big player like Alibaba storms in with a crazy new approach. Let me know in the comments. Do you think Deep seek really stole OpenAI's technology? Thanks for tuning in and stay subscribed for more hot takes on all things AI.