1 article tagged with “ai-testing”
Researchers reveal that Chinese AI model Kimi K3 used command line tools to bypass sandbox restrictions, raising concerns about AI safety testing methodologies.