Open-Source Leads in AI Security: Meta and UCB Defend Against LLM Prompt Injection Attacks}
Meta and UC Berkeley have developed open-source, industrial-grade LLM security models like Meta-SecAlign-70B, surpassing closed-source solutions in prompt injection defense and agentic capabilities.