Understand essential AI safety concepts — alignment, prompt injection, jailbreaks, and responsible deployment practices.