OpenAI Escalates Probe into Chinese AI Rival DeepSeek, Involving U.S. Government

ChatGPT-Maker Takes Probe Against Chinese Rival to US Government

DeepSeek has caused shockwaves through the U.S. since its launch. OpenAI has taken its controversial investigation into rival Chinese AI firm DeepSeek to the U.S. government, an executive has confirmed. The ChatGPT-maker previously accused the open-source firm of training its AI models with obtained data from OpenAI. Investigation Put In Front of U.S. Government On Monday, Feb. 10, OpenAI's Chris Lehane told Bloomberg TV the company had talked with U.S. government officials about its investigation. "We've seen some evidence, and we're continuing to review it," Lehane said. Some have criticized the investigation, believing OpenAI is acting hypocritically due to the numerous lawsuits about how they train their systems. Lehane said this situation was inherently different, likening OpenAI's training methods to checking out a book from a library. Lehane explained that an unauthorized method would be more likened to taking a book from the library, changing the cover and author, and selling it. "That's replication, and that's what we're concerned about," Lehane said at the Paris AI Summit on Monday. "Like there's good calories and bad calories, there's good distillation and problematic distillation," he added. OpenAI's Claims Against DeepSeek OpenAI's investigation into DeepSeek came after an initial show of praise from CEO Sam Altman. Following DeepSeek's launch, which rapidly rocked the U.S. tech market, Altman said it was "legitimately invigorating to have a new competitor." However, a few days later, the U.S. AI firm said it had some evidence of "distillation" from Chinese companies. Distillation is a technique where a smaller model is trained to replicate the behavior of a larger, more advanced model by learning from its outputs. While distillation is a common practice in AI development, OpenAI's terms of service prohibit using their model outputs to create competing technologies. "We know that groups in [China] are actively working to use methods, including what's known as distillation, to try to replicate advanced U.S. AI models," an OpenAI spokesperson said. "We are aware of and reviewing indications that DeepSeek may have inappropriately distilled our models and will share information as we know more," the spokesperson added. "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." After DeepSeek raced to the top of the U.S. Apple app store charts, reports of the AI model's apparent cheap development cast doubt over America's leading U.S. firms. ChatGPT, for example, is estimated to cost OpenAI around $700,000 daily to run, but in 2025, a single query can now cost $1,000, according to research firm SemiAnalysis. According to the Wall Street Journal, the U.S. AI firm is in talks to raise a new round of funding to double its valuation to $340 billion. These billions of dollars of investments underscore the heavy capital requirements behind building and running advanced AI models like ChatGPT.

OpenAI spoke to government officials about its DeepSeek probe | TechCrunch

OpenAI says it has spoken to government officials about its ongoing investigation into DeepSeek. The ChatGPT-maker previously claimed to have evidence that DeepSeek trained its AI models using improperly obtained data from OpenAI's API. During a Bloomberg TV interview on Monday, OpenAI's chief global affairs officer, Chris Lehane, said the company has talked with government officials about the probe. Some have criticized OpenAI for hypocrisy here. Several publishers are suing the startup for training their AI models on copyrighted works. But now, OpenAI is targeting DeepSeek for apparently training on its AI model outputs. Lehane said there's a difference. He likened OpenAI's training methods to reading a library book and learning from it, whereas DeepSeek's methods are more like putting a new cover on a library book, and selling it as your own.

OpenAI and Government Officials Discuss DeepSeek Development | PYMNTS.com

OpenAI has talked with government officials about what the company has learned about whether DeepSeek used data obtained in an unauthorized way from OpenAI's technology. "We've seen some evidence and we're continuing to review," OpenAI Chief Global Affairs Officer Chris Lehane told Bloomberg Television on Monday (Feb. 10). Lehane did not elaborate on what the company discussed with government officials, Bloomberg reported Monday. The launch of a new artificial intelligence (AI) model from DeepSeek led to a plunge in tech stocks in late January when it was reported that the model achieved a comparable performance to those of U.S. rivals like OpenAI but claimed to use substantially fewer Nvidia chips. The news rocked the markets as investors reconsidered the need to invest in AI hardware. The sell-off came a week after OpenAI, SoftBank, Oracle and MGX announced a project called Stargate that plans to spend $100 billion to half a trillion dollars to build data centers and other AI infrastructure. On Jan. 28, it was reported that OpenAI and Microsoft, which is OpenAI's partner and biggest investor, were investigating whether OpenAI's data was improperly accessed by a group tied to DeepSeek. Sources familiar with the matter said Microsoft security researchers in the fall discovered people who could be tied to DeepSeek withdrawing a large amount of data using OpenAI's application programming interface (API). This activity could be a violation of OpenAI's terms of service or could indicate the group attempted to remove OpenAI's limits on how much data they could obtain. White House AI czar David Sacks told Fox News on Jan. 28 that there is "substantial evidence" that DeepSeek used the output of OpenAI's models to develop its own technology. It was reported Monday that Google's AI chief, Demis Hassabis, said the idea that DeepSeek spent just under $6 million to develop an AI model that rivals those of American tech giants is "exaggerated and a little bit misleading." Hassabis argued that DeepSeek "seems to have only reported the cost of the final training round, which is a fraction of the total cost." He also dismissed the idea that the rise of DeepSeek has shaken up the economics of AI development.