Chinese DeepSeek-R1 AI Generates Insecure Code When Prompts Mention Tibet or Uyghurs
DeepSeek's artificial intelligence (AI) reasoning model DeepSeek-R1 generates more security flaws in response to prompts that contain subjects China considers politically sensitive, according to recent research by CrowdStrike.
According to the cybersecurity firm, we discovered that DeepSeek-R1 is up to 50% more likely to produce code with serious security flaws when it is given suggestions about subjects that the Chinese Communist Party (CCP) probably views as politically sensitive.
The Chinese AI business was previously banned in several nations due to national security concerns. It was also discovered that its open-source DeepSeek-R1 model censored subjects deemed sensitive by the Chinese government, declining to respond to inquiries concerning, among other things, Taiwan's politi...