Defenders now use prompt injection as a defensive tool

Ars Technica reports a shift where security defenders are embracing prompt injection techniques to protect AI systems. Instead of only attacking, prompt injection is being repurposed to harden LLMs against malicious commands.