Techniques and countermeasures to protect Large Language Models from specific attacks such as prompt injection, data poisoning, model inversion, and jailbreak. Includes input validation, output filtering, rate limiting, sandboxing, and behavioral monitoring strategies for production LLM systems.