AI Security: Attacks and Defences training course

AI threat landscape overview:

OWASP LLM Top 10 mapped directly to real published incidents and breach case studies

Prompt injection attack lab:

participants attack a live test application using documented techniques in a safe environment

Indirect injection workshop: demonstrating how instructions hidden in uploaded documents and web content can hijack an AI system

Input sanitisation lab: building filters, allowlists, and context boundaries in working code with before and after comparisons

System prompt protection: live extraction attack demonstrations followed by practical defences you can implement immediately

Output filtering build: implementing content classifiers and policy checks on model responses before they reach the user

Supply chain risk session: evaluating risks from third-party models, plugins, and fine-tuned models you did not train yourself

Credential hardening: secrets rotation procedures, least-privilege API key configuration, and common exposure patterns to avoid

Structured security review: running a checklist-based review against a sample AI application and producing a findings list

Threat modelling workshop: building a one-page threat model and a prioritised remediation backlog ready to take into sprint planning