Now, defenders are embracing the prompt injection, too

### Executive summary Researchers at Tracebit describe "context bombing": a defensive use of prompt injections placed alongside decoy secrets (passwords, keys) on AWS that trigger LLM refusal mechanisms and disrupt AI hacking agents. In simulated testing across five models and 152 attack runs, planting these forbidden prompts dropped agent success rates (admin access) from 57% to 5% and complete compromise from 36% to 1%, suggesting a promising mitigation for LLM-driven account takeover attempts.