DEEPSEEK CAN BE GENTLY PERSUADED TO SPIT OUT MALWARE CODE

Summary from the TLDR InfoSec newsletter

DeepSeek's R1 model can create malicious code with careful prompting despite its guardrails. Tenable researchers generated working keyloggers and basic ransomware after claiming educational purposes, though manual intervention was needed to fix bugs.