US Agencies: Six Chinese AI Firms Stole Data from OpenAI, Google

US cybersecurity agencies including the NSA, CISA, and FBI have issued a joint advisory warning that six Chinese AI firms systematically stole billions of data tokens from American models since late 2024. The report identifies companies like DeepSeek, Moonshot AI, Alibaba, and MiniMax as using advanced tricks to bypass safety limits and extract knowledge from leaders such as OpenAI and Google. Officials believe these industrial-scale operations likely had Chinese government knowledge to accelerate their own model development.

The agencies describe this as a planned national-level action that undermines fair competition and US technological leadership. Between late 2024 and mid-2025, firms allegedly extracted data from models like GPT-5 and Claude to train their own systems. While model distillation is a normal practice, these companies reportedly violated terms of service to steal proprietary reasoning capabilities, giving them a significant advantage in training advanced AI systems.

Senator Bernie Sanders is set to host a closed-door briefing for Congress next Wednesday to discuss AI advancements, featuring prominent researchers. This event aligns with his long-standing advocacy for increased investment in AI research and legislation to promote the field. Meanwhile, Andrew Tulloch, a star AI researcher at Meta, is leaving the company less than a year after joining, despite receiving a pay package worth more than $1 billion.

Outside of security concerns, generative AI is reshaping online shopping by allowing customers to talk to search bars instead of typing keywords. Recent data shows AI-driven sessions on Shopify have doubled with higher conversion rates. In transportation, Pony.ai CEO James Peng predicts robotaxis will be common on streets within five years, with the firm already operating in Beijing and Shanghai. T-Mobile is also leveraging AI to reroute network capacity during disasters like hurricanes to keep customers connected.

Key Takeaways

  • US agencies warn six Chinese AI firms stole billions of tokens from American models since late 2024.
  • DeepSeek, Moonshot AI, Alibaba, and MiniMax are named in the report for bypassing safety limits.
  • Chinese firms allegedly extracted data from OpenAI and Google models like GPT-5 and Claude.
  • Officials suspect the Chinese government knew about these operations to speed up model development.
  • Senator Bernie Sanders will host a closed-door AI briefing for Congress next Wednesday.
  • Meta AI researcher Andrew Tulloch is leaving after receiving a $1 billion+ pay package.
  • Generative AI on Shopify has doubled sessions and improved conversion rates for shoppers.
  • Pony.ai CEO predicts robotaxis will be common on streets within five years.
  • T-Mobile uses AI to reroute network capacity during disasters like hurricanes and wildfires.
  • New StereoPolicy technology gives robots 3D vision without depth sensors using standard cameras.

US Agencies Warn Chinese Firms Stole AI Data

US cybersecurity agencies say six Chinese AI companies stole billions of data tokens from American models since late 2024. The firms, including DeepSeek and Moonshot AI, used advanced tricks to bypass safety limits and extract knowledge from leaders like OpenAI and Google. Officials believe the Chinese government likely knew about these industrial-scale operations to speed up their own model development. The report suggests these actions threaten US technological leadership and recommends better detection systems to stop the theft.

China AI Firms Systematically Steal US Tech Secrets

The NSA, CISA, and FBI released a report stating that Chinese AI companies are systematically stealing capabilities from US frontier models. Between late 2024 and mid-2025, firms like DeepSeek and Moonshot extracted data from models such as GPT-5 and Claude to train their own systems. The agencies describe this as a planned national-level action that undermines fair competition and US technological leadership. They recommend defensive measures like behavioral monitoring and aggressive responses to high-confidence malicious requests.

US Accuses Chinese Companies of Stealing AI Models

US government agencies claim Chinese AI firms covertly extracted billions of tokens from major US models to reduce their own development costs. The joint advisory from the NSA and CISA names companies like Alibaba and MiniMax for using evasive techniques to access data from OpenAI and Google. While model distillation is a normal practice, these firms allegedly violated terms of service to steal proprietary reasoning capabilities. The report warns this strategy gives Chinese companies a significant advantage in training their own advanced AI systems.

Senator Sanders to Host AI Briefing for Congress

Senator Bernie Sanders will host a briefing on artificial intelligence for his Senate colleagues next Wednesday. The event will feature prominent researchers to discuss the latest advancements and their impact on fields like healthcare and transportation. Sanders has long advocated for increased investment in AI research and has introduced legislation to promote the field. The briefing is expected to be a closed-door session limited to senators and their staff.

Illustration Shows Deepseek Logo and AI Words

An illustration displays the Deepseek logo alongside the words 'Artificial Intelligence AI'. The image was taken on January 29, 2025. It serves as a visual representation of the company's focus on artificial intelligence technology.

Generative AI Changes How Online Shopping Works

Generative AI is changing online shopping by allowing customers to talk to search bars instead of typing keywords. Shoppers can now describe needs like 'lightweight black running shoes under $100' and get personalized results. This shift moves product discovery from simple keyword matching to understanding context and user intent. Recent data shows AI-driven sessions on Shopify have doubled, with higher conversion rates than traditional search.

Robotaxis Could Change How Cities Function Soon

James Peng, CEO of Pony.ai, predicts robotaxis will be common on streets within five years. The Chinese firm already operates services in major cities like Beijing and Shanghai and is expanding to Europe and South Korea. Widespread adoption could save parking space and change urban planning since private cars are used for only two hours a day. Other companies like WeRide and Baidu are also pushing robotaxi services globally.

T-Mobile Uses AI to Improve Disaster Response

T-Mobile is using artificial intelligence to make its network faster and smarter during disasters like hurricanes and wildfires. CTO John Saw explained that the AI system reroutes network capacity to areas that need it most when emergencies strike. This technology helps customers stay connected when they need it most during critical times. The company is proud to use AI to make a real difference in disaster response efforts.

Star AI Researcher Andrew Tulloch Leaves Meta

Andrew Tulloch, a star AI researcher at Meta, is leaving the company less than a year after joining. He was offered a pay package worth more than $1 billion when he was hired. The departure marks a significant change for Meta's AI research team. Details about his next move have not been fully disclosed yet.

New Tech Gives Robots 3D Vision Without Sensors

Researchers have created a new method called StereoPolicy that gives robots 3D vision without using depth sensors. The system uses pairs of standard cameras to understand spatial relationships and grasp objects in cluttered environments. Tests show this method succeeds 59% of the time, significantly better than traditional single-camera or depth-map approaches. This breakthrough helps robots perform complex tasks like hanging glass cups without failing on transparent objects.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

AI Data Theft Chinese AI Companies US Cybersecurity Agencies AI Model Extraction National-Level AI Action AI Model Distillation AI Research Briefing AI in Online Shopping Robotaxis AI in Disaster Response AI Researcher Departure Robotic 3D Vision

Comments

Loading...