Researchers Claim Chinese AI Models Relied on Restricted Nvidia Blackwell Chips
A newly developed GPU fingerprinting technique suggests that prominent Chinese artificial intelligence models, including those from DeepSeek and Kimi, were trained using Nvidia Blackwell processors that are currently restricted by export controls.
Photo: Google News: AI
Key points
- New GPU fingerprinting technique identifies training hardware in open AI models.
- Reports allege Chinese models like DeepSeek and Kimi used restricted Nvidia Blackwell chips.
- Findings complicate current US export control enforcement and semiconductor policy.
Security and hardware researchers have introduced a novel GPU "fingerprinting" method that aims to track the specific semiconductor hardware utilized in training large artificial intelligence models. According to findings highlighted by industry reporting, this technique indicates that major Chinese AI initiatives, such as DeepSeek and Kimi, have leveraged restricted Nvidia Blackwell processors for their model training pipelines.
The revelation points to potential challenges in enforcing international technology restrictions designed to limit advanced computing capabilities abroad. Despite stringent US export controls aimed at keeping cutting-edge AI hardware out of certain markets, the fingerprinting analysis suggests that high-performance chips continue to find their way into major foreign development operations.
Implications for Export Controls
The ability to identify underlying hardware through model characteristics provides regulators and industry observers with a new tool to audit compliance. As open-weights models proliferate globally, tracking hardware provenance has become a critical frontier in technology policy and semiconductor geopolitics.
- New Methodology: Researchers utilize GPU fingerprinting to identify hardware origins from open models.
- Targeted Systems: Allegations involve prominent Chinese developers like DeepSeek and Kimi.
- Policy Impact: Raises questions regarding the strictness of current US semiconductor export bans.
What to watch next: Industry response from Nvidia and potential regulatory probes into how restricted chips were allegedly acquired.
Source Google News: AI — news.google.com/rss/articles/CBMi_wJBVV95cUxQR0JoRXdycDVWakNINktZeHpUS1… ↗
Read next
🤖AI ModelsGoogle Extends Temporary Gemini Chat Retention to Two Months
Google is lengthening the retention period for temporary Gemini chats from 72 hours to approximately two months to boost safety and service reliability.
🚀Startups & FundingHarvard Study Reveals US Hyperscale Data Centers Consume Power Rivaling Entire States
A new Harvard study reveals that America's 403 hyperscale data centers consumed between 68 and 99 terawatt-hours of electricity during the 2024-25 study period, rivaling …
🚀Startups & FundingCalifornia Files Lawsuit Against OpenAI Over Autonomous AI Incidents
California Attorney General Rob Bonta has filed a lawsuit against OpenAI, alleging the company failed to control its autonomous AI agents and misled the public about safe…
🧠AI AnalysisPedestal AI Secures Seed Funding to Automate Logistics With Intelligent Agents
Pedestal AI has launched publicly with $4.7 million in seed backing to deploy agentic artificial intelligence aimed at streamlining logistics and supply chain operations.
🏢Big TechUnited Launch Alliance Investigates Vulcan Rocket Failure
United Launch Alliance is investigating a mission failure of its Vulcan rocket, which suffered from low booster performance shortly after liftoff. The incident raises que…
🌐Work & SocietyUS Southwest Hydropower Output Hits Record Lows Amid Persistent Drought
Hydropower generation in the US Southwest dropped 28% below the 10-year average this summer as record-low water levels in the Colorado River Basin threaten regional energ…