Large Language Models (LLMs) such as ChatGPT, Claude, Gemini, and Bard are revolutionizing the way we create, communicate, and learn. Once confined to research labs, they are now integrated into education platforms, finance, healthcare, and customer support. However, as adoption grows, so do the risks. Attackers may exploit vulnerabilities through prompt injections, data poisoning, or model inversion to recover sensitive information. This course introduces the foundations of LLM security, emphasizing both technical understanding and ethical awareness.

Over the span of this course, you will build a strong understanding of both offensive and defensive dimensions of LLM security. Learners will explore adversarial prompts that bypass safeguards, data poisoning that embeds malicious outputs, and model inversion attacks that extract private data. To counter these threats, the course provides a toolkit of defense methods, including robust training, prompt sanitization, adversarial testing, and privacy-preserving frameworks such as differential privacy and secure computation.

As generative AI expands into classrooms, hospitals, and creative industries, the importance of secure and ethical deployment cannot be overstated. Unprotected models risk spreading misinformation, leaking private data, or reinforcing bias. Beyond technical defenses, this course integrates legal and ethical perspectives from the EU AI Act, UNESCO guidelines, and the NIST AI Risk Management Framework. By completion, learners will possess the skills to detect, mitigate, and critically evaluate LLM vulnerabilities, preparing them to responsibly shape the future of artificial intelligence.