OpenAI says DeepSeek may have used its AI outputs 'inappropriately' to train new models

• OpenAI investigates DeepSeek for potentially using its AI outputs inappropriately.
• DeepSeek built AI models using less-advanced chips at a fraction of the cost of US rivals.
• "We take aggressive, proactive countermeasures to protect our technology," OpenAI said.

OpenAI is probing whether DeepSeek inappropriately trained its powerful AI models using the US startup's technology.

A spokesperson for OpenAI said the company is reviewing the matter closely and would take "aggressive, proactive countermeasures" to protect its AI models from improper use.

"We know that groups in the PRC (People's Republic of China) are actively working to use methods, including what's known as distillation, to try to replicate advanced US AI models," the spokesperson wrote in an email. "We are aware of and reviewing indications that DeepSeek may have inappropriately distilled our models, and will share information as we know more."

DeepSeek built top-performing AI models using less-advanced chips and for what it says is a fraction of the cost of rivals such as OpenAI, Google, and Meta. Some tech stocks have been hammered by the news this week, which put a big question mark over massive spending on AI chips and related infrastructure.

It's a particular challenge to OpenAI because DeepSeek models are priced way below the US startup's offerings.

OpenAI lets developers with a valid license integrate its proprietary models into their own applications. Its terms of use, however, prohibit developers from using outputs from its models to develop any models that directly compete with its products and services.

David Sacks, the White House's artificial intelligence and crypto czar, told Fox News "it's possible" that DeepSeek engaged in intellectual property theft from OpenAI.

"We take aggressive, proactive countermeasures to protect our technology and will continue working closely with the U.S. government to protect the most capable models being built here," the OpenAI spokesperson told BI on Wednesday.

Microsoft notified OpenAI that its security researchers in the fall had observed individuals they believed may be affiliated with DeepSeek siphoning a large amount of data using OpenAI's application programming interface, or API, Bloomberg reported on Tuesday.