LLM binary vulnerability detection, AI evaluation ecosystems, adversarial maritime defences, red-teaming scaling laws, and collaborative patching with smaller models.