Day 1: Understanding Generative AI Attack Surfaces

Application Architecture and Assets

• Performance signals around application architecture
• Containment steps before wider spread
• Privilege changes needing closer review
• Attack paths across trust boundaries

Prompt and Data Threats

• Application limits around data threats
• Access anomalies needing prompt investigation
• Recovery actions after security events
• Threat signals requiring rapid triage

Model and Supply Chain Exposure

• Decision points for supply chain
• Adversarial tests prior to production release
• Detection gaps exposed by testing
• Identity risks linked to context

Day 2: Designing Security Controls for AI Applications

Input and Output Guardrails

• Ownership expectations for output guardrails
• Escalation thresholds for high-impact threats
• Evidence preserved for incident review
• Malicious patterns across security telemetry

Access, Secrets, and Isolation

• Operational risks around access and secrets
• Interface contracts between system components
• Integration failures needing fallback paths
• Dependencies affecting service reliability

Retrieval and Tool Security

• Validation checks for tool security
• Trust boundaries around sensitive data
• Data flows across connected services
• Monitoring signals after production release

Day 3: Testing and Operating AI Security

Adversarial Testing and Monitoring

• Performance signals around adversarial testing
• Service dependencies mapped prior to change
• Capacity limits shaping system performance
• System components and integration boundaries

Incident Response and Recovery

• Practical tests for incident response
• Failure recovery for critical components
• Operational telemetry informing diagnosis
• Security controls embedded by design

Generative AI Security Standard

• Decision points for generative AI
• Evidence retention for later oversight
• Exceptions handled with documented rationale
• Policy limits limiting unauthorized use

• Decision simulation testing data threats in context
• Hands-on workshop applying access and secrets to realistic tasks
• Peer review challenging assumptions around adversarial testing
• Practical lab developing generative AI for operational use