OpenAI, Microsoft, Trump admin claim DeepSeek trained AI off stolen data

DeepSeek is causing havoc throughout the AI industry. U.S.-based tech companies that have heavily invested in AI saw their stocks take a tumble this week after the China-based startup released a new AI model on par with OpenAI’s latest model, yet much cheaper to train — plus, DeepSeek made it free and open source.

Now, OpenAI and Microsoft are hitting back at DeepSeek. The companies claim they have evidence that the Chinese company trained their AI model by using data it had lifted from OpenAI.

OpenAI claimed to the Financial Times that it had evidence DeepSeek trained its AI by using OpenAI’s models. According to Bloomberg, Microsoft also believes that an OpenAI developer account it believes is connected to DeepSeek stole large amounts of data late last year. Microsoft is a major investor in OpenAI, providing the company with billions of dollars in resources.

OpenAI’s Sam Altman previously said that the company spent more than $100 million to train GPT-4. DeepSeek says it trained its latest model DeepSeek-Ri with less than $6 million.

In an interview with Fox News, President Trump’s AI and crypto czar David Sacks backed those claims, saying, “There’s substantial evidence that what DeepSeek did here is distilled the knowledge out of OpenAI’s models.”

Sacks said that it’s “possible” that DeepSeek engaged in IP theft.

As multiple outlets have pointed out, there is a sense of irony in the situation that OpenAI has found itself in with DeepSeek.

OpenAI itself has been largely criticized for training its AI models off of, well, anything it could suck up on the internet. Many YouTube creators were shocked to find out that OpenAI used their content without consent in order to create its products.

YouTuber Marques Brownlee, for example, pointed out recently how OpenAI’s video generation model, Sora, created a desk with a plant, just like the plant visible in his own videos, when he requested a tech review video be generated. 

Based on OpenAI’s statement to Bloomberg, the company is likely to take action in an attempt to protect their IP.

“As the leading builder of AI, we engage in countermeasures to protect our IP, including a careful process for which frontier capabilities to include in released models, and believe as we go forward that it is critically important that we are working closely with the U.S. government to best protect the most capable models from efforts by adversaries and competitors to take U.S. technology,” OpenAI said.

Will content creators be able to follow suit and use the same argument to protect their IP from OpenAI? We’ll see how this unfolds.