The AI industry is buzzing with controversy after OpenAI and U.S. officials raised concerns that Chinese AI startup DeepSeek may have copied ChatGPT’s technology. While no concrete evidence of intellectual property (IP) theft has been disclosed, the allegations have sparked intense debate.
- Founded by Liang Wenfeng in May 2023, DeepSeek is a Chinese AI startup that has gained attention for its open-source approach.
- The company is funded by High-Flyer, a hedge fund also run by Wenfeng, allowing it to grow rapidly without external investors.
- DeepSeek has successfully built AI models using older chips due to U.S. export bans on high-performance Nvidia chips.
- Released last week, R1 is DeepSeek’s flagship reasoning model, which performs at or above OpenAI’s o1 model on various benchmarks.
- Built on V3 and based on Alibaba’s Qwen and Meta’s Llama, R1 is open-source, meaning anyone can download and use it.
- However, DeepSeek has not disclosed its training dataset.
- DeepSeek offers significantly lower pricing than U.S. competitors, with R1 API access starting at $0.14 per million tokens, compared to OpenAI’s $7.50 per million tokens.
Key Takeaways:
DeepSeek Accused of Distilling OpenAI's Models
- OpenAI and U.S. President Donald Trump’s AI adviser, David Sacks, suggest DeepSeek leveraged a technique known as "distillation" to extract knowledge from OpenAI’s models.
- OpenAI’s terms explicitly prohibit distillation, and it is actively banning accounts attempting to use this method.
DeepSeek’s Model Showed Signs of ChatGPT Influence
- Users noticed that DeepSeek’s AI sometimes identified itself as ChatGPT or referenced OpenAI’s terms and policies.
- Experts believe DeepSeek may have trained its model using vast amounts of ChatGPT-generated interactions.
DeepSeek’s AI Model Costs Raise Eyebrows
- The company claims to have trained its flagship model for just $5.6 million, far lower than the billions spent on OpenAI’s ChatGPT.
- Analysts question how DeepSeek achieved such cost efficiency, especially given U.S. restrictions on high-performance AI chips.
Privacy Concerns Around DeepSeek
- Similar to concerns raised around TikTok, DeepSeek’s privacy policy states that user data may be stored on servers in China.
- The policy specifies that DeepSeek collects various types of data, including:
- IP address, unique device identifiers, cookies
- Date of birth, username, email, phone number, and password
- Text or audio input, chat history, feedback, and uploaded files
- Proof of identity or age when required
- While the company claims compliance with data protection laws, it does not mention GDPR compliance.
Broader Implications for the AI Industry
- OpenAI has been accused of similar practices—using copyrighted materials from publishers without permission.
- This raises questions about ethical AI development and whether big tech companies should be the ones calling out potential IP violations.
- The U.S. government is working with OpenAI to prevent adversaries from replicating American AI advancements.
What’s Next?
DeepSeek’s rapid AI advancements have unsettled markets, but without hard evidence of IP theft, the debate remains speculative. This case highlights the growing AI arms race and the blurred lines between innovation, competition, and potential infringement.
?? What are your thoughts? Should AI companies be more transparent about their training data? Drop your comments below!
