Safe, reliable AI: detect hallucinations, verify answers, design better prompts, and add simple guardrails.