Microsoft thinks China’s DeepSeek might have used OpenAI’s tech before its big breakthrough
In This Story
Microsoft (MSFT+2.86%) and OpenAI are investigating whether a group connected to the Chinese AI startup DeepSeek accessed OpenAI’s data without permission, according to a new report.
Bloomberg, citing unnamed sources familiar with the matter, reports that Microsoft’s security team noticed unusual activity last fall in which individuals, believed to be linked to DeepSeek, accessed large amounts of data through OpenAI’s application programming interface (API).
The ChatGPT maker’s API allows software developers to license and use the company’s proprietary AI models into their own applications. However, the breach may have violated OpenAI’s terms of service and appeared to be an attempt to bypass restrictions on how much data can be accessed from the service. Microsoft, OpenAI’s largest investor, flagged the issue to the company, Bloomberg reports.
Last week, the Hangzhou-based AI startup DeepSeek introduced its first-generation, open-source reasoning models, DeepSeek-R1, that demonstrated comparable performance to OpenAI’s reasoning models.
The release prompted a global sell-off of tech stocks, with the Nasdaq, Dow Jones Industrial Average, and S&P 500 all falling Monday morning. Chipmaker Nvidia (NVDA+7.29%) saw its stock plummet 17%, losing close to $600 billion in value — the largest single-day market cap loss ever for a U.S. company.
At the same time, the mobile app for DeepSeek’s AI chatbot, also called DeepSeek, surged to the top of Apple’s (AAPL+3.88%) App Store, while its site experienced outages from an influx of new users.
Investors were spooked by the Chinese startup, which released its DeepSeek-V3 model in December that it said cost just $5.6 million to train and develop on Nvidia’s reduced-capability H800 chips. U.S. rivals such as OpenAI and Meta have touted spending tens of billions on Nvidia’s more powerful chips.
The U.S. National Security Council is already reviewing the potential national security risks associated with DeepSeek, White House Press Secretary Karoline Leavitt told reporters Tuesday. Leavitt noted that President Donald Trump views the startup’s newly released AI model as a “wake-up call” for American developers.
—Britney Nguyen contributed to this article.
