Tech giant Microsoft has initiated a thorough investigation into DeepSeek, a Chinese AI company, over suspicions that it may have utilized OpenAI's application programming interface (API) for the development of its artificial intelligence models. This probe follows closely on the heels of claims made by David Sacks, described as Trump’s AI and crypto “czar,” who publicly stated that the Chinese company used OpenAI’s models to train its own models. Security researchers affiliated with Microsoft have indicated that there is a possibility that DeepSeek copied substantial amounts of data from OpenAI’s API, a concern that Microsoft had previously communicated to OpenAI, a company in which it is a major investor.

OpenAI's terms of use explicitly forbid the use of its API output to develop competing AI models, stating, “You are prohibited from using Output to develop models that compete with OpenAI.” The terms also restrict users from automatically or programmatically extracting data or output. In an interview, David Sacks reiterated that there is “substantial evidence” suggesting DeepSeek “distilled” knowledge from OpenAI’s AI models, a process he characterized as theft. While Sacks did not disclose the source of this evidence, he asserted that DeepSeek employed responses generated by OpenAI models to train its proprietary AI. The allegations suggest potential legal ramifications if DeepSeek circumvented OpenAI’s rate limits to query its API at scale, utilizing a method commonly referred to as distillation, where one model extracts knowledge from another in a teacher-student dynamic.

The swift rise of DeepSeek, particularly with the recent launch of its open-source R1 model and R1 chatbot, has significantly intensified pressures within the tech industry. The R1 model's functionality and accuracy, comparable to its U.S. counterparts despite using fewer resources and less computing power, have been lauded as a significant advancement for the AI industry as a whole. Its R1 chatbot quickly soared to the top of global app download charts over the weekend, reflecting its widespread appeal. Nvidia recognized DeepSeek's work as an "excellent AI advancement" and a prime example of "Test Time Scaling," noting its ability to create new models using widely available resources and export-control compliant computing, which still requires substantial Nvidia GPUs for inference.

DeepSeek's streamlined and cost-efficient approach to AI development stands in stark contrast to the capital-intensive strategies prevalent in Silicon Valley. This disruption has had notable financial repercussions; Nvidia, a major chip-making company, saw its stock slip significantly, losing nearly $600 billion from its market capitalization in a single trading day. This market shift also impacted several of the world's wealthiest individuals, including Oracle co-founder Larry Ellison and Nvidia Corp.’s co-founder Jensen Huang, who collectively lost an estimated $108 billion. The startup’s cutting-edge large language models (LLMs) and disruptive pricing strategy have reshaped the AI landscape, drawing comparisons to leading players like OpenAI.

Beyond the market impact, DeepSeek has also come under scrutiny from U.S. government officials. The National Security Council is currently reviewing the implications of DeepSeek’s applications, which have notably topped the Apple App Store. Furthermore, the U.S. Navy has imposed a ban on the use of DeepSeek’s AI, citing "potential security and ethical concerns." DeepSeek was founded in May 2023 by Liang Wenfeng, a prominent figure in both the hedge fund and AI sectors, and first entered the competitive AI market with the release of DeepSeek Coder in November 2023.