Photo Credit: Village Global (Sam Altman)
DeepSeek, a Chinese AI research lab, has released an advanced AI model which rivals leading models from OpenAI. The DeepSeek-R1 model can perform complicated mathematical reasoning, code generation, and more with fewer resources than its American competitors.
DeepSeek originated as Fire-Flyer, a deep-learning research branch of High-Flyer, a Chinese quantitative hedge fund first founded in 2015. High-Flyer first gained recognition by using advanced computing to analyze financial data. In 2023, the fund’s founder Liang Wenfeng pivoted toward AI research and established DeepSeek to develop new AI models.
DeepSeek-R1 employs reinforcement learning techniques and multi-stage training to enhance its capabilities. The company has open-sourced its flagship model along with six smaller variants under an MIT license—giving developers the ability to refine and commercialize the models freely. Now OpenAI is questioning how DeepSeek is able to deliver its AI model that outpaces OpenAI’s model using inferior hardware.
Both Bloomberg and the Financial Times report that Microsoft and OpenAI are probing whether the Chinese company trained the R1 model on outputs of OpenAI models. “Such activity could violate OpenAI’s terms of service or could indicate the group acted to remove OpenAI’s restrictions on how much data they could obtain,” the Bloomberg report reads.
Meanwhile, ‘AI Czar’ David Sacks says he believes there is ‘substantial evidence’ that DeepSeek was able to “distill the knowledge out of OpenAI’s models.”
“There’s a technique in AI called distillation, which you’re going to hear a lot about, and it’s when one model learns from another model. Effectively what happens is that the student model asks the parent model a lot of questions, just like a human would learn. But Ais can do this asking millions of questions and essentially mimic the reasoning process they learn from the parent model kind of sucking the knowledge of the parent model,” Sacks said in a Fox News interview.
“There’s substantial evidence that what DeepSeek did here is they distilled the knowledge out of OpenAI’s models and I don’t think OpenAI is very happy about this.” OpenAI is currently fighting several lawsuits that allege the company gobbled large amounts of data from the internet in an ‘unauthorized manner,’ which was in many cases, a violation of terms of service from those they took the data from.
“It is (relatively) easy to copy something that you know works,” OpenAI CEO Sam Altman tweeted in December 2024. “It is extremely hard to do something new, risky, and difficult when you don’t know if it will work. It’s also extremely hard to rally a big talented research team to charge a new hill in the fog together. This is the key to driving progress forward.”
OpenAI has argued that it will protect its intellectual property, which was built on top of stolen IP from millions of other sources. “We know PRC based companies—and others—are constantly trying to distill the models of leading US AI companies,” an OpenAI spokesperson told Bloomberg.
“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 US government to best protect the most capable models from efforts by adversaries and competitors to take US technology.”
Essentially, OpenAI is arguing that it is legal for the company to scrape the internet for resources to train—no matter if those resources are illegal or not. So scraping pirate sites, Z-library, torrents, and other illegal sources of data is a-okay—but knowledge distillation—a common learning strategy among LLMs is not.