Risk Managing The DeepSeek Disruption and Commoditization of AI

Never has an industry been so starkly split between whether to use a technology or not. But wait… the same thing happened in 2023 with ChatGPT.

Some days after the announcement of Trump’s USD500 billion Stargate project which I was tempted to go through with a fine-tooth comb, the Internet blew up with how a Chinese lab’s AI model, DeepSeek-R1, was unbelievably low cost compared to Open AI’s own ChatGPT.

DeepSeek is a Hangzhou-based AI company founded in 2023. The buzz around DeepSeek began to grow after the release of its R1 model on January 20, 2024, but it was only after the launch of its AI assistant, R1, on January 10, 2025, that the world stood up and listened.

Then two days before Chinese New Year, tech investors dumped stocks evaporating USD593 billion of NVIDIA’s market value (about USD1 trillion when totaled with other AI tech companies’ stocks) and CXposè.tech editor Leona Lo decided to put her CNY celebration plans on hold. By then there was so much hype, and she spent a good part of the festive season wading through all the conjecture to discover real gems.

Our snapshot recaps the main trains of thought about DeepSeek now, a majority of which are circulating around Linkedin.

Like ChatGPT’s o1, DeepSeek’s R1 is a “reasoning” model. When you query ChatGPT, you may note how it explains or gives the reasoning behind its answers. DeepSeek-R1 works similarly, which may explain why ChatGPT was overtaken as the number one free AI chat app on Apple’s App Store.

Geopolitical concerns

Being based in China, DeepSeek challenges US technological dominance in AI.

Countries that have banned it like Italy and Australia are not satisfied with its (lack of) data protection measures, and reports are circulating about how it is training its model on all kinds of user data. Ironically, countless individuals are also sighing out loud, “If only DeepSeek’s data didn’t reside in China on Chinese servers.”

Those that want more visibility can audit the actual source code and compile the LLM for their own use themselves.

When ChatGPT came around last year, similar debates were going on. Same as with ChatGPT, there are several camps for and against Deepseek, and they are all vociferously vocal about why to use it or not.

Former Government CISO Chai Chin Loon observed as much that the risks surrounding use of LLMs or large language models have not changed.

Chai Chin Loon, a former government CISO in Singapore

In response to news reports about countries banning or potentially banning China’s Deepseek technology, he opined, “Frankly, I view that much of the public controversies around geopolitical risks are being played up more than anything else.”

Any organization, be it from the private or public sector, needs to go back to doing proper risk assessment – what does it want to protect at what cost, how much is it prepared to pay, and what would the impact be if they didn’t and there is a loss, he explained.

The geopolitical risk aspect diminishes pressing technology concerns regarding LLMs that need to be overcome. DeepSeek can actually help to address these concerns because of its decision to be open source and transparent about their research. Those that want more visibility can audit the actual source code and compile the LLM for their own use themselves.

According to the World Economic Forum (WEF), “DeepSeek is open source, meaning it is available for anyone to download, copy and build upon. Its code and comprehensive technical explanations are freely shared, enabling global developers and organizations to access, modify and implement.”

Snapshot

U.S. restrictions on China’s AI advancements – they don’t seem to work

This technological leap has reignited debates about the effectiveness of U.S. restrictions on China’s AI advancements. And it’s not just AI. Before DeepSeek trended, there were reports of very advanced robots sighted in China.

In any case, the achievement of this China-made AI assistant raised questions about how DeepSeek achieved such remarkable cost-efficiency in AI development.

The export of the highest-performance AI accelerator and GPU chips from the US is restricted to China. Despite that, DeepSeek has demonstrated that leading-edge AI development is possible without access to advanced US technology, and better yet, at about 1/7th the cost it took to develop US’s finest.

Let’s also not forget how Huawei went ahead and developed their own mobile operating system, Harmony OS, which was launched within months of Huawei being placed on a US trade blacklist that barred American firms without license from selling tech and software to it.

Cost disruption

DeepSeek claims to have developed its R1 model for less than $6 million. At a glance, this low-cost development threatens the business model of US tech companies that have invested billions in AI. DeepSeek is also cheaper to use than OpenAI’s ChatGPT.

Business model threat

In contrast with OpenAI, which is proprietary technology, DeepSeek is open source and free, challenging the revenue model of US companies charging monthly fees for AI services.

Companies and developers find workarounds for DeepSeek

With what has been shared above, we shouldn’t be surprised that businesses want to tap into some of that DeepSeek goodness. There are news reports that Meta is reverse engineering it, and tech companies like Microsoft, Amazon, and NVIDIA are rushing to adopt the R1 model in their solutions.

Security shortcomings

From outdated jailbreak defenses to a real data leak, DeepSeek has demonstrated significant security shortcomings. Theori.io, a cybersecurity firm with Korean roots, published on its blog, “Its rapid deployment outpaced security hardening, leaving known exploits open and sensitive data unprotected.

This exposes users to risks like malicious output, data theft, or misuse of the AI for harmful purposes.

“Robust security testing and proactive patch management appear to be areas where DeepSeek lags, especially when measured against industry best practices.” the blog stated.

At the Core of the DeepSeek Drama

After peeling aside most of the layers obfuscating the DeepSeek debacle, Chin Loon opined, “It is a new innovation that disrupts the whole cost calculation of LLM, and that fundamentally affects investors more than anybody else, because the returns on investments (ROI) calculation suddenly changed significantly.”

Chin Loon said that once this has been established and the furore dies down, then the real work of technology evaluation comes in. “There is still a need to control what users put in, what the data is trained on, whether the LLM should be on-premise, on the cloud, or take a hybrid approach, as well as consider other vulnerabilities like prompt hacks, hallucinations, and so on.

“Technology risk is just one of many risks that an organization handles. My view is that geopolitical risk is something that most companies don’t have to deal with and they should instead prioritize focusing on risks to the company’s interests.

“Ultimately, it boils down to the work by the CISOs, the IT guys, the system owners… it really has to boil down to the risk assessment (of using such a technology).”

The next AI frontier: Interface loyalty

Enough about models and LLMs, when are AI tech companies going to pay attention to interface loyalty? Greg Isenberg, raises a good point about interface loyalty, in one of his many Linkedin posts.

This CEO of a company that oversees a portfolio of Internet companies, identified a main hurdle for AI startups, and said, “Getting millions of people to make your product part of their daily workflow – that’s the real barrier to entry. ChatGPT didn’t win because it had the best model. It won because it was dead simple to use. And I think it has staying power because of that.”

Greg Isenberg

He also observed that AI startups are actually competing on being the default way humans will interact with AI and opined that while technology advantage is temporary, “Interface lock-in is forever.”

The way forward?

China is going to go ahead and do what it wants to do irrespective of hurdles the US throws in its direction. Regardless, Deepseek did one very major and important thing for the tech world when it showed how it very ably and competently did a whole lot more, with less. Significantly it may even have democratized the tech know-how and capability needed to build LLMs.

AI is gradually being commoditized and the arrival of DeepSeek is a signal that it is fast approaching. As the industry grapples with this shift, the focus may well move from who can build the most powerful AI to who can integrate it most seamlessly into daily life. DeepSeek’s achievement suggests that the future of AI lies not in exclusivity, but in AI being ubiquitous.