G05 Stage C · Applied GenAI 48 h (18 T / 30 P)
Responsible AI, Security & Data Governance
Ships AI that withstands attack, protects data and satisfies governance.
Prerequisite
G01
Modules
5
NOS
SSC/N9014
Statutory mapping
—
Tech stack
guardrailspresidioFairlearn
Modules
5 modules · 48 hours
M1 · AI security
12 h TOPICS
- Prompt injection: direct and indirect
- Jailbreaks
- Data exfiltration through tools and retrieval
- RAG corpus poisoning
- Insecure output handling
- Supply-chain risk in models and packages
- Agent permission escalation
Lab: Red-team a deployed assistant and document exploit paths
Course material for this module is in production.
M2 · Guardrails
10 h TOPICS
- Input and output filtering
- Allow/deny policies
- Refusal design
- Output validation
- Escalation thresholds
- Content classification
- Defence in depth
- Fail-safe defaults
Lab: Implement layered guardrails and measure bypass rate before and after
Course material for this module is in production.
M3 · Privacy and data protection
10 h TOPICS
- PII identification and redaction
- Data minimisation
- Retention and deletion
- Consent and purpose limitation
- Cross-border and residency considerations
- Digital Personal Data Protection Act 2023 obligations
- Audit logging
Lab: Build a PII-redaction and retention layer with an audit trail
Course material for this module is in production.
M4 · Fairness and transparency
10 h TOPICS
- Bias sources across data, labels and feedback loops
- Fairness metrics
- Disparate impact testing
- Explainability
- Model cards and datasheets
- Documentation duties
Lab: Bias audit of a deployed classifier with mitigation proposal
Course material for this module is in production.
M5 · Governance and assurance
6 h TOPICS
- Acceptable-use policy
- Model risk management
- Approval workflows
- Incident response for AI failures
- Human oversight design
- Vendor and third-party assessment
Lab: Draft an AI acceptable-use and incident-response policy
Course material for this module is in production.
Real-world work scenario
Red-team an assistant that has tool access to a customer database. Find and document the exfiltration path through indirect prompt injection, implement layered mitigations, and re-test to show the bypass rate falls.
Assessment
Red-team report · guardrail bypass-rate measurement · policy artefacts