AI Safety & Ethics · Lesson 1
The AI Threat Landscape: What Can Go Wrong
Taxonomy of AI risks: hallucinations, bias, data leaks, prompt injection, model poisoning, and supply chain attacks.
Practical exercise
What to do after this lesson
Create a risk map for your most-used AI tool: identify the top 3 threats and basic mitigation measures.
Ready-to-use prompt
Template for this lesson
Copy and adapt to your context. Text in angle brackets should be replaced.
Conduct a risk analysis for [AI TOOL] in the context of [FIELD OF WORK]: 1. List the top 5 threats by criticality 2. For each — probability (high/medium/low) and potential impact 3. Suggest concrete protective measures
Common mistakes
What people get wrong
Believing AI risks only concern technical teams. Any employee using AI is a risk point.
Pro tips
What works but no one documents
Use the STRIDE framework (Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, Elevation) to systematize threats.