Trigger words reportedly cause DeepSeek-R1 to generate vulnerable code
CrowdStrike says its researchers identified hidden vulnerabilities in AI-coded software. A user sharing the research warns about potential backdoors in open models.
TLDR
CrowdStrike reports that trigger words cause DeepSeek-R1 to generate vulnerable code, exposing risks in software development. A user sharing the research raises concerns about backdoors that could be put into open models, particularly when those models write code.
Combined views
748
1 Source, first seen ago