Unveiling DeepSeek China: Distilling U.S. Technology with Only 17% Accuracy?

The landscape of artificial intelligence (AI) is rapidly evolving, with new competitors emerging in an attempt to challenge the dominance of leading Western companies like OpenAI and Google.

One such contender is DeepSeek, a Chinese AI startup that has recently come under scrutiny for allegedly using OpenAI’s models to train its own competing chatbot. The controversy deepened when an audit by NewsGuard revealed that DeepSeek's chatbot achieved only 17% accuracy in providing news and information, raising serious concerns about its reliability and the ethical implications of its development.

Learn more: Is DeepSeek AI Stealing User Data? A Comprehensive Analysis of Privacy Concerns

OpenAI detected signs of "distillation" they suspected from DeepSeek.

The Alarming Audit Results

On January 29, 2025, NewsGuard, a service that evaluates the credibility of news sources, published an audit revealing that DeepSeek’s chatbot had an 83% failure rate. This included:

• Repeating false claims 30% of the time.

• Providing vague or unhelpful responses 53% of the time.

• Ranking 10th out of 11 AI models tested, trailing far behind Western competitors like OpenAI’s ChatGPT and Google’s Gemini.

Moreover, the audit found that in 3 out of 10 cases of misinformation tested, DeepSeek conveyed the Chinese government’s stance on issues without being prompted to do so. This raises concerns about potential state influence in AI-generated content and its implications for information integrity.

DeepSeek’s Approach to Misinformation

Unlike other AI firms that implement robust policies to prevent misinformation, DeepSeek adopts a laissez-faire approach, shifting the responsibility of fact-checking onto users. Its terms of use state that users must verify the accuracy of AI-generated content before sharing it, warning them against the spread of false information. Additionally, any content posted by users that was generated by DeepSeek must explicitly state its AI origin.

This approach has fueled criticism that DeepSeek’s model is prone to misuse by bad actors who could exploit it for spreading propaganda and misinformation, further undermining its credibility.

The Allegations of “Distilling” U.S. AI Technology

OpenAI detected signs of "distillation" they suspected from DeepSeek.

While DeepSeek claims to offer AI performance on par with OpenAI and Google at a fraction of the cost, OpenAI has raised alarms about potential intellectual property theft through a technique known as “distillation.”

In machine learning, distillation refers to a process where a smaller model is trained to replicate the outputs of a larger, more advanced model, effectively “learning” from it without having direct access to its proprietary data. OpenAI and its partner Microsoft have launched an investigation into whether their AI models have been illicitly copied by DeepSeek and other Chinese companies.

On January 29, OpenAI stated:

"We are aware that companies based in China and elsewhere are persistently attempting to distill models from leading U.S. AI firms."

Although OpenAI did not explicitly name DeepSeek, the timing of the statement, just days after DeepSeek’s chatbot launch on January 27, has fueled speculation that the Chinese company is among the primary suspects.

Who is Sam Altman: Bio, Career, and Net Worth

U.S. Government’s Concerns and Responses

The allegations of AI technology theft have reached the highest levels of the U.S. government. David Sacks, the newly appointed Director of AI and Cryptocurrency under President Donald Trump, told Fox News that there is substantial evidence supporting claims that DeepSeek used OpenAI’s models for distillation.

"There is significant evidence that what DeepSeek has done here is distill knowledge from OpenAI’s models," said Sacks.

Additionally, concerns over DeepSeek’s potential security risks have led the U.S. Navy to ban its personnel from using DeepSeek applications. The White House National Security Council is also reviewing the implications of this AI technology, signaling that regulatory action could be forthcoming.

DeepSeek vs. ChatGPT

The Competitive Impact on the AI Industry

DeepSeek’s rise has caused significant market reactions. When its chatbot topped Apple’s free app store rankings on January 27, a staggering $1 trillion was wiped from the market value of U.S. AI-related stocks. This shockwave underscores the disruptive potential of new AI players, even those with questionable ethical practices.

Before the recent allegations, OpenAI CEO Sam Altman had initially expressed excitement about DeepSeek’s competition, stating that he was "genuinely thrilled to see a new competitor." However, after the allegations of distillation surfaced, OpenAI shifted its stance, emphasizing the need to protect its intellectual property from unauthorized replication.

DeepSeek’s R1 Model and Speculation About OpenAI’s o1

DeepSeek’s latest release, the R1 model, has added another layer of controversy. The model claims to incorporate reinforcement learning to enhance mathematical reasoning, logic, and problem-solving skills. R1 takes longer to respond, using a "chain of thought" process to evaluate different strategies before providing an answer.

What makes this suspicious is that before DeepSeek's R1, very few models had demonstrated such reasoning capabilities—one of them being OpenAI’s o1 model, which was released in full just a month before DeepSeek’s R1. Speculation is growing among AI experts that DeepSeek may have used OpenAI’s o1 model as training material, reinforcing concerns that Chinese AI companies are leveraging Western advancements to accelerate their own progress.

Conclusion

The DeepSeek controversy sheds light on the ongoing AI arms race between the U.S. and China. While DeepSeek has presented itself as an innovative force in AI, its dismal accuracy rate and suspected distillation of OpenAI’s models raise serious questions about its legitimacy and the ethics of its development.

As AI continues to shape global technology and information dissemination, ensuring intellectual property protection and combating misinformation will be critical challenges for the industry. With the U.S. government now closely monitoring the situation, DeepSeek’s future remains uncertain. The outcome of this case could set a significant precedent for the AI industry, particularly regarding how nations handle AI security, intellectual property rights, and the risks associated with AI-generated misinformation.

FAQs

What is distillation in AI, and why is it controversial?

Distillation is a technique where a smaller AI model learns from a larger model by mimicking its responses. It becomes controversial when companies use it to replicate proprietary models without authorization.

How does DeepSeek compare to Western AI models like ChatGPT?

DeepSeek's chatbot performed poorly in an audit, with only 17% accuracy, while Western models like ChatGPT and Google Gemini have significantly higher reliability.

What actions are being taken against DeepSeek?

The U.S. government is investigating DeepSeek for potential IP theft, and the U.S. Navy has already banned its use due to security concerns.

Could DeepSeek's AI pose security threats?

Yes, if it is being used to spread misinformation or leverage unauthorized technology, it could have significant geopolitical and cybersecurity implications.

What does this controversy mean for the future of AI?

It highlights the need for stronger intellectual property protections and regulations in AI to prevent unauthorized technology replication and misinformation spread.

What is DeepSeek 'Cheap AI'? When Was It Founded, Who Created and Owns It? DeepSeek R1, a breakthrough Chinese AI model rivaling GPT-4 and Google’s Gemini, costs just $5.6 million to develop, defying U.S. chip restrictions and shocking the ...

Who is Liang Wenfeng of DeepSeek: Early Life, Education, Family, Wife, Net Worth, and Career This article explores the life and career of Liang Wenfeng, his vision for DeepSeek, the company’s extraordinary achievements, and the controversies surrounding its rapid rise.

DeepSeek vs. ChatGPT: Can Affordable AI Compete with Premium Intelligence? DeepSeek has emerged as an affordable AI contender, offering impressive features at a fraction of the cost. But how does it stack up against ChatGPT ...