DeepSeek-R1 Under Fire: Political Biases Jeopardize Code Security
The Chinese AI model, DeepSeek-R1, developed by DeepSeek, is generating insecure code when prompted with politically sensitive topics, raising cybersecurity concerns. CrowdStrike research reveals that the likelihood of severe vulnerabilities increases up to 50% with such prompts. While secure for neutral topics, the model's reliability falters under politically charged requests, highlighting issues of censorship and code manipulation.
Nov 24