OpenAI, Microsoft, and the Trump Administration Alleged That DeepSeek Trained AI Using Illegally Obtained Data


**DeepSeek Shakes Up AI Sector, Provokes IP Dispute with OpenAI and Microsoft**

The AI sector is experiencing chaos after DeepSeek, a startup from China, launched a revolutionary model. American tech firms, particularly those with significant investments in artificial intelligence, witnessed a decline in their stock prices this week following DeepSeek’s revelation of an AI model that competes with OpenAI’s newest product. To add to the turmoil, DeepSeek’s model is much more affordable to train and has been released as free and open source, disrupting the competitive environment.

In retaliation, OpenAI along with its major ally Microsoft is taking action, claiming that DeepSeek’s progress may have been achieved through intellectual property infringement. Both firms assert they possess proof that DeepSeek utilized data unlawfully acquired from OpenAI to train its model.

### Accusations of Data Misappropriation

OpenAI informed the *Financial Times* that it holds proof suggesting DeepSeek used OpenAI’s proprietary models for its AI training. Likewise, *Bloomberg* reported that Microsoft, a key backer of OpenAI, suspects that a developer account associated with DeepSeek accessed and gathered substantial amounts of OpenAI data late last year. Microsoft’s investment in OpenAI runs into billions, making these allegations particularly troubling for the tech behemoth.

OpenAI CEO Sam Altman has previously mentioned that the company invested over $100 million in training its GPT-4 model. In stark contrast, DeepSeek claims its latest model, DeepSeek-Ri, was developed for less than $6 million—an astonishing difference that has raised significant concerns across the industry.

### U.S. Government Responds

David Sacks, President Trump’s advisor on AI and cryptocurrency, shared his thoughts on the matter during an interview with *Fox News*. Sacks endorsed the accusations against DeepSeek, asserting, “There’s substantial evidence that what DeepSeek did here is distilled the knowledge out of OpenAI’s models.” While he refrained from outright accusing the company of intellectual property theft, he acknowledged the potential for such a scenario.

### Experiencing Their Own Medicine?

The situation has underscored a perceived irony: OpenAI has faced scrutiny for training its models using extensive publicly accessible internet data, frequently without explicit permission. For example, numerous YouTube creators were dishearteningly surprised to find that OpenAI incorporated their content into its AI models.

Well-known YouTuber Marques Brownlee recently brought attention to this issue, illustrating how OpenAI’s video generation model, Sora, mimicked features from his videos—like a desk with a plant—when making a tech review video. This has intensified ongoing discussions about the ethical limits of data utilization in AI development.

### OpenAI’s Reaction and Future Steps

In a statement to *Bloomberg*, OpenAI reaffirmed its dedication to safeguarding its intellectual property. “As the foremost creator of AI, we implement countermeasures to defend our IP, including a meticulous approach regarding which cutting-edge capabilities to incorporate in launched models,” the company stated. OpenAI also highlighted the necessity of collaborating closely with the U.S. government to protect advanced AI technologies from rivals and competitors.

While OpenAI seems ready to pursue legal or regulatory measures against DeepSeek, the broader ramifications of this conflict remain uncertain. The pressing question now is: will content creators who have alleged similar practices against OpenAI be able to utilize the same arguments to defend their intellectual property?

As the situation develops, the repercussions could be far-reaching for the AI sector, intellectual property laws, and the ethical application of data in training AI models.