Now, defenders are embracing the prompt injection, too

Prompt injection has long been an attacker's tool against LLMs, but defenders are now adopting the technique for their own purposes. The article explores how context bombing and guardrails are being used to protect AI platforms.