NIST Report Finds Security Flaws in Chinese DeepSeek AI Models
The report found DeepSeek models were 12 times more likely to be hijacked and had a 94% jailbreak rate compared to U.S. models.
Significant Security Shortcomings
A new report from the U.S. National Institute of Standards and Technology (NIST), released on Sep 30, 2025, detailed significant security shortcomings in models from Chinese AI developer DeepSeek. The evaluation found that AI agents built on DeepSeek's models were, on average, 12 times more likely to be hijacked by malicious instructions than the U.S. frontier models that were tested.
The report serves as a strong counterpoint to the idea that performance on capability benchmarks is the only metric that matters. It highlights that security and alignment can vary widely between models from different developers and …
Archive Access
This article is older than 24 hours. Create a free account to access our 7-day archive.