DeepSeek Splashes Onto the Tech MarketDeepSeek Splashes Onto the Tech Market
China’s recently released artificial intelligence (AI) platform is challenging US-based AI. But it may not be safe.
February 1, 2025

At a Glance
- China’s AI has been in development for years, long before ChatGPT showed up.
- DeepSeek R1 exhibited a 100% attack success rate, meaning it failed to block a single harmful prompt.
- DeepSeek showed up with a model that rivals GPT-4o1 but costs a fraction to run.
On January 10, 2025, DeepSeek released its first free chatbot app. The DeepSeek-R1 model was designed for iOS and Android. Two weeks after its introduction, DeepSeek-R1 surpassed ChatGPT as the most-downloaded free app on the iOS App Store in the US. The attention caused Nvidia's share price to drop by 18%.
DeepSeek is owned and funded by Chinese hedge fund High-Flyer. Its co-founder, Liang Wenfeng, established the company in 2023 and serves as its CEO. The app is an open-source AI model that uses a Mixture-of-Experts approach, making it more efficient and cost-effective than ChatGPT. It excels in technical tasks and mathematical computations, while ChatGPT offers better user experience and broader capabilities.
Hitting the tech market like a hand grenade
The appearance of DeepSeek caught US investors by surprise. Yet China’s AI has been in development for years, long before ChatGPT showed up. “DeepSeek might seem like it came out of nowhere, but this has been in the works for a while. China’s AI ecosystem has been quietly evolving,” Tinglong Dai from Johns Hopkins University – and a member of the Institute for Operations Research and the Management Sciences (INFORMS) – told Design News. “Before ChatGPT launched in 2022, China was already a world leader in real-world applications of vision AI – think facial recognition, autonomous logistics, and large-scale surveillance. In many ways, China was ahead of the US, and there was a real sense of anxiety here about losing the AI race.”
When ChatGPT was released by OpenAI, it caused a storm. Much like the Dot Com boom, OpenAI created a flurry of activity and investment. “ChatGPT changed everything. It kicked off the generative AI era and gave the US a clear edge,” said Dai. “Silicon Valley took the lead, and Washington stepped in to make sure it stayed that way, chiefly through export controls on advanced microchips. For a while, it seemed like that strategy was working.”
While DeepSeek seemed like a surprise. The US tech world long expected China to deliver stiff AI competition. “China has been adapting AI for a long time. DeepSeek is proof that AI isn’t just about brute-force compute anymore – it’s about efficiency,” said Dai. “The old assumption was that cutting-edge AI required billions of dollars, massive GPU clusters, and access to the best chips. DeepSeek just shattered that myth. Investors were betting on AI being a scarce, tightly controlled resource. Now they’re realizing it might not be scarce at all.”
Cisco evaluates DeepSeek for its security
Researchers from Robust Intelligence – a part of Cisco and the University of Pennsylvania – completed an evaluation of DeepSeek to determine its level of security. The team used algorithmic jailbreaking techniques, applying an automated attack methodology on DeepSeek R1. They tested it against 50 random prompts from the HarmBench dataset. “These covered six categories of harmful behaviors including cybercrime, misinformation, illegal activities, and general harm,” noted the summary of the report. “The results were alarming: DeepSeek R1 exhibited a 100% attack success rate, meaning it failed to block a single harmful prompt. This contrasts starkly with other leading models, which demonstrated at least partial resistance.”
The Robust Intelligence findings suggest that DeepSeek’s claim of cost-efficient training methods, including reinforcement learning, chain-of-thought self-evaluation, and distillation may have compromised its safety mechanisms. “Compared to other frontier models, DeepSeek R1 lacks robust guardrails, making it highly susceptible to algorithmic jailbreaking and potential misuse,” noted the report.
Investors are stuck with US AI
The stock selloff in response to the appearance of DeepSeek suggests investors are suddenly less enamored by US-based AI. Yet it also seems they have nowhere else to go. “The real shock isn’t DeepSeek itself; it’s what it represents,” said Dai. “For years, the AI industry operated on the assumption that only a handful of companies could control access to top-tier models. Then DeepSeek shows up with a model that rivals GPT-4o1 but costs a fraction to run. That’s a paradigm shift.”
Investors will have to rethink AI in the face of a less expensive model. “It means AI isn’t just for the tech giants anymore. New players will emerge, and the big players will have a much harder time defending their moats,” said Dai. “The old business model – where AI is expensive, centralized, and locked behind API paywalls – just took a serious hit.”
DeepSeek’s tech market explosion
Even with the paradigm shift, the leading AI companies still hold a commanding position in the market. “Leaders like OpenAI and Anthropic are not necessarily threatened. AI isn’t just about raw performance – it’s about trust, reliability, and ecosystem,” said Dai. “Remember the early days of search engines? There were plenty of competitors, but Google won – and keeps winning – because even small advantages in accuracy and integration matter. Same with AI. Businesses want something they can trust and depend on. OpenAI still has that edge.”
In the early days of search where the offerings were varied, Google figured out a way to monetize a free service. The world of AI monetization is still in its early days. “The AI business model is shifting fast. AI is getting cheaper, faster, and harder to control. Investors who thought a handful of companies would dominate forever now have to rethink everything,” said Dai. “We’re moving toward a world where top-tier AI doesn’t just live in the cloud – it runs locally, on personal devices, and is available to anyone who wants it. That’s a massive shift, and it’s only just beginning.”
AI into the future
So, where are we going with AI? We asked Dai to look into his crystal ball. “As they say, it’s hard to predict – especially when it comes to the future. But one thing’s clear: AI isn’t slowing down,” said Dai. “It’s just getting real. The hype cycle is fading, and we’re moving past the phase where AI is a shiny new toy. It’s becoming infrastructure – something you don’t think about but rely on every day.”
Ultimately – like the internet – tech companies are working to serve up apps and infrastructure that we can’t do without. “The companies that figure out how to make AI useful – not just bigger or more expensive – are the ones that will win,” said Dai. “The real shift isn’t in who builds the best model; it’s in who integrates AI so seamlessly that you stop noticing it. AI will be everywhere, like air – always there, shaping industries, decisions, and everyday life, but mostly invisible.”
About the Author
You May Also Like