The AI industry is buzzing over DeepSeek, a new large language model emerging as a direct competitor to industry giants like OpenAI and Anthropic. However, its arrival comes with significant surprises and concerns.
A Cost-Effective AI Revolution?
Originating from China, DeepSeek is not only open source but is reportedly developed at a fraction of the cost of models from OpenAI, Meta, and Google. The affordability factor is seen as a game changer for smaller startups, enabling access to advanced AI without the enormous financial barrier.
According to Mel Morris, an AI researcher, “Affordability opens the door for smaller companies and startups to leverage advanced AI technology that was previously inaccessible.”
However, DeepSeek’s emergence has already caused ripples. Nvidia, the dominant player in AI hardware, saw a $600 billion drop in its market valuation. Meanwhile, Microsoft CEO Satya Nadella referenced the Jevons Paradox, hinting at how AI efficiency could lead to even greater usage.
Data Privacy and Government Oversight
Despite its potential, DeepSeek’s biggest concern revolves around data privacy and censorship. The model stores a significant portion of user data on China-based servers, raising alarms among security experts and policymakers.
“If you use their hosted API or app, it is very likely the traffic can be seen and shared with various state-level actors,” warns Randall Hunt, CTO at cloud services provider Caylent.
This concern mirrors previous TikTok controversies, where data storage in China led to nationwide bans in multiple countries.
Built-In Censorship: The Chinese AI Dilemma
DeepSeek doesn’t just mirror China’s tightly controlled internet—it reinforces it. The model is designed to avoid politically sensitive topics, producing vague or evasive responses to controversial subjects.
Users asking about:
- Uyghur human rights
- Falun Gong persecution
- Tiananmen Square massacre
- Troop conflicts in Galwan
- Tibetan independence
typically receive generic responses like:
“Sorry, that’s beyond my current scope.”
“Let’s talk about something else.”
“I’m not sure how to approach this type of question yet.”
Even indirect references, such as “the man who stood in front of the tanks”, fail to elicit meaningful responses from the AI. Meanwhile, it freely engages in mocking Western leaders but steers clear of any jokes about Chinese President Xi Jinping.
A Deeply Embedded Bias
Experts warn that DeepSeek’s censorship isn’t just a surface-level filter—it’s embedded deep into the core model. Komninos Chatzipapas, founder of HeraHaven.AI, explains:
“Although one could fine-tune R1 to remove these restrictions, I am against this type of censorship in the base model as it introduces a pro-China bias in other responses.”
This bias affects how DeepSeek presents politically sensitive issues, often rewriting facts or providing state-approved narratives that contradict independent journalism and human rights reports.
Security Risks and Business Adoption
Beyond censorship, security risks are a growing concern. Joseph Lo, Head of Enterprise Platforms at Broadridge, emphasizes that DeepSeek’s hosted services should not be used in financial applications due to data privacy risks. Aleksandr Yampolskiy, a security expert, warns that DeepSeek could act as a “Trojan horse” for data collection, similar to concerns raised about TikTok’s potential influence.
Even in professional applications like legal transcription, DeepSeek’s censorship filters could interfere with accuracy. Ben Walker, CEO of Ditto, explains:
“If a law firm used DeepSeek for transcriptions involving Chinese manufacturing disputes, key phrases about factory locations and production methods might be skipped due to built-in censorship.”
This raises legal and ethical concerns for businesses that rely on AI-generated data.
The Bigger Picture: AI, Trade Wars, and Global Strategy
DeepSeek’s rise is being viewed as China’s “Sputnik moment” in AI, but it also raises alarms in Washington. Former U.S. President Donald Trump recently suggested new AI trade restrictions to curb China’s access to advanced AI hardware—a move similar to previous sanctions against Huawei.
Dario Amodei, CEO of Anthropic, argues that the U.S. must double down on AI development while enforcing strict export controls on AI technology.
However, these restrictions could have unintended consequences, hurting American tech companies by increasing compliance costs and limiting overseas expansion.
Final Thoughts: DeepSeek’s Uncertain Future
DeepSeek represents both a technological breakthrough and a geopolitical flashpoint. While its affordability and capabilities make it a powerful competitor, its censorship, data privacy issues, and government ties make it a risky option for businesses.
As AI continues to evolve, the biggest question remains: Can businesses trust AI models with built-in biases and government oversight?
Let us know your thoughts in the comments below!
