Cybersecurity & AI Security Specialist

Module 1 · Overview of AI Risks · Lesson 1 of 1

Module overview — Overview of AI Risks

Module overview — Overview of AI Risks Goal : Map the attack surfaces specific to AI and LLM systems, from user input down to the logs. Study time : Durée d'étude : 4 h Prerequisites : Notions de réseau et de sécurité (le cours Cybersécurité est recommandé en amont) Running thread : Wanjiru, in Nairobi. Wanjiru is the security officer of a distribution company in Nairobi. She must review the company's customer assistant, which is connected to a document base and to internal tools. Her usual checklist, designed for web applications, says nothing about prompts, documents or tools. In this module she learns to draw the map of the places where this system can be attacked. Lesson plan : 1. Why classic security is no longer enough 2. OWASP Top 10 for LLM map 3. Anatomy of a vulnerable AI stack What you will be able to do : - Explain how an LLM blurs the boundary between instructions and data - Match each risk of the OWASP list to a real component of an AI architecture - Spot at least six entry points in a toy stack and rank their priority End-of-module questions (Answer in writing, with figures.) : 1. How many distinct entry points did you spot in the toy stack? Give a number. 2. How many risks from the OWASP list could you attach to a specific component? Give a figure. 3. What are the two questions that guide the reading of an AI architecture? State them. Module deliverables : - Your list of the entry points of the CasaVolt stack, with their priorities - Your architecture sheet of an AI assistant, filed in your notebook

Free preview, no account needed — the rest of this module and the following modules unlock after enrolling.