OpenAI has accused Chinese artificial intelligence start-up DeepSeek of utilizing its proprietary AI models to train a competing open-source system, raising concerns about potential intellectual property (IP) theft. The allegation, first reported by the Financial Times, has intensified debates over the protection of cutting-edge AI research and technology. In response, Western governments are now considering imposing strict sanctions on China’s AI sector.
The creators of ChatGPT claim they have uncovered evidence of “distillation,” a technique where smaller AI models learn from the outputs of larger, more advanced ones. This method enables companies to replicate high-performance models at a fraction of the cost and computational resources.
DeepSeek’s Rapid Rise Raises Questions
DeepSeek’s recently released R1 reasoning model has stunned industry observers with its cognitive capabilities, achieving performance comparable to leading US AI models. However, suspicions arose when experts noted that DeepSeek’s model displayed response patterns similar to OpenAI’s GPT-4, suggesting it may have been trained using OpenAI’s outputs—an act that would breach OpenAI’s terms of service.
OpenAI declined to provide detailed evidence but reiterated its policy that prohibits users from copying or utilizing its outputs to develop competing models. A company insider emphasized that while distillation is an accepted industry practice, using it to create a rival product is problematic.
DeepSeek has yet to respond to requests for comment.
US Officials Weigh In on Alleged IP Theft
David Sacks, the White House AI and cryptocurrency advisor under President Donald Trump, has echoed OpenAI’s concerns. Speaking to Fox News, Sacks stated that “it is possible” DeepSeek engaged in IP theft through distillation.
“There’s a technique in AI called distillation… when one model learns from another model and kind of sucks the knowledge out of the parent model,” said Sacks. “And there’s substantial evidence that what DeepSeek did here is they distilled the knowledge out of OpenAI models, and I don’t think OpenAI is very happy about this.”
Despite these claims, Sacks did not present direct proof to support his allegations.
West Considers Sanctions on Chinese AI
As concerns over China’s AI development escalate, Western nations are now exploring punitive measures to curb Beijing’s rapid advancements in artificial intelligence. A coalition of US and European lawmakers is reportedly pushing for economic sanctions targeting Chinese AI firms accused of intellectual property theft.
“There must be serious consequences for those who engage in the unauthorized appropriation of cutting-edge AI technology,” said a US State Department official. “We are considering a range of sanctions that could include export restrictions, financial penalties, and limitations on access to Western semiconductor technology.”
These potential measures come amid a broader technological standoff between China and the West, with Washington already imposing restrictions on high-end chip exports to Chinese firms. A move to impose additional sanctions could further strain relations and escalate the ongoing tech war.
AI Distillation: An Industry-Wide Practice?
AI experts acknowledge that using outputs from leading AI models to train smaller ones is common. Ritwik Gupta, a PhD candidate in AI at the University of California, Berkeley, noted that many AI developers—both in China and the US—leverage human-aligned commercial LLMs like ChatGPT for their training processes.
“It is a very common practice for start-ups and academics to use outputs from human-aligned commercial LLMs to train another model,” Gupta explained. “That means you get this human feedback step for free. It is not surprising to me that DeepSeek supposedly would be doing the same.”
However, OpenAI’s case against DeepSeek highlights a broader concern: the financial and strategic challenges frontier AI companies face in protecting their technological advantages while others capitalize on their work.
China’s AI Industry Gains Momentum
Chinese tech firms have rapidly advanced in AI research, often using innovative approaches to maximize efficiency with limited computational resources. DeepSeek reportedly used just 2,048 Nvidia H800 graphics cards and $5.6 million to train its V3 model, which boasts 671 billion parameters—a fraction of what OpenAI and Google spent on models of similar scale.
OpenAI claims that Chinese companies are continually seeking ways to extract knowledge from leading US AI firms. The company has implemented countermeasures to safeguard its IP, including selective deployment of frontier capabilities in public models.
“We know [China]-based companies—and others—are constantly trying to distil the models of leading US AI companies,” OpenAI said in a statement. “We engage in countermeasures to protect our IP… and believe it is critically important that we work closely with the US government to best protect the most capable models from efforts by adversaries and competitors to take US technology.”
OpenAI Faces Copyright Allegations of Its Own
Ironically, OpenAI is currently battling legal claims regarding its own alleged misuse of copyrighted content. Media outlets, including The New York Times, and prominent authors have sued the company, accusing it of training its AI models on their content without permission.
