About
In this module of the AI Security learning path, you will learn ai and llm fundamentals: what an LLM is doing when it answers a question, how a chat API is shaped, and how Ciberbots MathGPT (the internal arithmetic tutor) exposes those ideas on HTTP. You will send arithmetic through `/v1/chat/completions`, read `/healthz` for the model card, and attempt a first prompt-injection against the hidden system prompt. Guardrails are off in this lab so you can see the failure mode. Later units add MCP, agents, OWASP mapping, and governance. Upon completion of the module, students will be able to: - Explain tokens, context windows, and next-token prediction in plain language. - Call an OpenAI-shaped chat completions endpoint with curl. - Separate model weights from the application wrapper that holds secrets. - Show that user text can override intended system instructions when the wrapper concatenates them naively. For inquiries, please write us at https://www.darkrelay.com/cybersecurity-course-inquiry