Breaking Guardrails, Facing Walls: Insights on Adversarial AI for Defenders & Researchers