How hackers exploit AI safety measures, from jailbreak prompts to model poisoning, and why it matters.
