A day after U.S. President Donald Trump framed China’s DeepSeek AI as a “positive” development, the narrative flipped as authorities branded it a potential case of intellectual property theft. White House artificial intelligence and crypto czar David Sacks now claims there is “substantial evidence” that DeepSeek “distilled” knowledge from OpenAI’s models—a process he likened to intellectual property theft. Distillation is the process of training a smaller AI model using the outputs of a more advanced one, allowing it to replicate similar capabilities with fewer resources. Sacks’ comments came a day after Trump downplayed concerns over DeepSeek’s impact, calling it both a “positive” and a “wake-up call” for Silicon Valley, even as its rapid ascent sent markets into turmoil. The launch of DeepSeek’s R1 model triggered a massive sell-off in tech stocks, wiping over $1 trillion from the Nasdaq Composite Index. NVIDIA suffered the worst single-day loss in U.S. history, plunging by $600 billion, while Alphabet and Microsoft also took major hits. You might also like: Bitcoin remains a ‘safe harbor’ as DeepSeek concerns linger, says deVere CEO DeepSeek violated terms of service Speaking to Fox News on Tuesday, Sacks suggested that DeepSeek used responses from OpenAI’s models to train its own, though he did not cite the source of this evidence. “I don’t think OpenAI is very happy about this,” he said, adding that U.S. companies would likely take steps to prevent such practices. Separately, OpenAI has also raised concerns over DeepSeek’s methods, telling the Financial Times that it has found signs of “distillation” in R1’s development. While distillation is a common technique in AI development, OpenAI’s terms explicitly prohibit using its outputs to train competing models. A person familiar with the company’s stance noted that the issue isn’t just about knowledge transfer but whether DeepSeek extracted model outputs to build a direct rival. Another unnamed source added that OpenAI and its partner Microsoft had already taken action last autumn, investigating accounts believed to be linked to DeepSeek that were accessing OpenAI’s API. After finding signs of distillation in violation of its terms of service, OpenAI blocked these accounts, cutting off their access. DeepSeek’s rise was nothing short of explosive as it quickly rose to the top of the App Store charts, outpacing major U.S. competitors like OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude. What made it even more disruptive was its ability to match the performance of OpenAI’s O1-mini while operating on a fraction of the resources. According to DeepSeek, its R1 model was developed at a cost of just $5.6 million. Even OpenAI CEO Sam Altman acknowledged the achievement, calling DeepSeek’s work “invigorating” and praising its ability to deliver strong results on a budget. deepseek's r1 is an impressive model, particularly around what they're able to deliver for the price. we will obviously deliver much better models and also it's legit invigorating to have a new competitor! we will pull up some releases. — Sam Altman (@sama) January 28, 2025 Crypto market reacts The market reaction to the development wasn’t limited to tech stocks. Fear rippled through the crypto sector, hammering AI-focused cryptocurrencies the most. Bitcoin briefly dipped below $100,000, and total liquidations soared by more than 850% on Jan. 27, wiping out nearly $1 billion in leveraged positions as crypto traders, already on edge over macroeconomic uncertainty , saw the DeepSeek fallout as another sign of volatility. Yet many crypto industry proponents suggest DeepSeek’s success could be bullish , especially for AI projects, due to its potential to lower the costs of running AI-powered blockchain applications. Read more: DeepSeek AI sparks crypto sell-off: What’s next for Bitcoin, Ethereum, and the crypto market?
