Insider Brief
- U.S. technical experts are investigating whether individuals linked to Chinese AI startup DeepSeek improperly accessed OpenAI data, raising concerns over intellectual property theft and AI security.
- Microsoft researchers observed unauthorized data transfers from OpenAI’s API by individuals allegedly connected to DeepSeek, prompting OpenAI to strengthen its protective measures and cooperate with the U.S. government.
- DeepSeek’s AI assistant has overtaken ChatGPT on Apple’s U.S. App Store, sparking a tech stock selloff and national security concerns, with the U.S. Navy and Australian officials warning against its use.
Technical experts in the United States are investigating whether a group allegedly linked to Chinese artificial intelligence startup DeepSeek improperly obtained data from OpenAI, according to reports from several media sources, including Asia Financial and Reuters.
Microsoft researchers observed what they described as an unauthorized data transfer, or exfiltration, involving individuals believed to be connected to DeepSeek, according to sources. The data was accessed through OpenAI’s application programming interface (API), which developers and businesses use to integrate OpenAI’s models into their own applications. The incident reportedly occurred in the last quarter of 2024.
Suspected Data Theft Raises Security Concerns
Microsoft, OpenAI’s largest investor, alerted the company to the suspicious activity. While Microsoft declined to comment on the findings, OpenAI issued a statement noting that China-based firms and other foreign competitors frequently attempt to replicate the models of U.S. AI leaders.
“We engage in counter-measures 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,” an OpenAI spokesperson said, as reported in Asia Financial.
David Sacks, the White House’s AI and crypto czar, suggested in an interview with Fox News that DeepSeek may have stolen OpenAI’s intellectual property, according to Asia Financial
“There’s substantial evidence that what DeepSeek did here is they distilled the knowledge out of OpenAI’s models,” Sacks said.
DeepSeek could not immediately be reached for comment.
DeepSeek’s AI Model Shakes Up the Market
The allegations surfaced as DeepSeek’s AI assistant surged in popularity, briefly overtaking OpenAI’s ChatGPT on Apple’s U.S. App Store. The company, which positions itself as a low-cost alternative to major American AI firms, triggered a tech stock selloff earlier this week.
The sudden success of DeepSeek’s AI model has raised broader industry concerns about the cost of large-scale AI development. Some analysts argue that U.S. tech giants may be spending too much on massive data centers to sustain their AI ambitions.
“DeepSeek makes it very clear that the current trajectory of scaling up of data centers is highly unlikely to be economic to Nvidia’s customers,” said Gil Luria, head of tech research at investment group DA Davidson, in an interview with CNN.
The OpenAI-backed Stargate project, which aims to build a network of AI supercomputing facilities, represents one of the largest bets on AI infrastructure. If companies like DeepSeek can produce competitive AI models with fewer resources, some industry observers say it could reshape the economics of AI development.
National Security Implications Under Review
Beyond the economic implications, U.S. officials are evaluating potential security risks tied to DeepSeek’s model. The White House’s National Security Council is reviewing the situation, but according to a report from Axios, the precise national security risks remain unclear.
The U.S. Navy, however, has already issued guidance warning personnel not to use DeepSeek’s AI model for work or personal tasks, citing “potential security and ethical concerns associated with the model’s origin and usage,” according to CNBC.
In Australia, cybersecurity executives and federal opposition leaders have also raised alarms. The Sydney Morning Herald quoted cybersecurity expert Alistair MacGibbon: “Anything we enter into DeepSeek is going straight into the hands of the Chinese Communist Party.”
DeepSeek Responds to Attacks
Amid the scrutiny, DeepSeek briefly limited new user registrations on Monday, citing “large-scale malicious attacks” on its platform. The company later resumed operations but has not publicly addressed the allegations of data exfiltration.
This case will likely not be going away anytime soon, experts suggest, adding that it will continue to fuel ongoing debates over AI security, data protection, and economic strategy. While U.S. tech firms continue to push the limits of AI infrastructure, DeepSeek’s rise — and the questions surrounding it — underscore the challenges of protecting advanced AI models in an increasingly competitive and complex landscape.
