Course Format and Delivery

Delivery Method: LiveOnline
Schedule: 2 sessions of 4.5 hours
Cost: $1,695 USD

All sessions are delivered live by an expert instructor in a fully interactive online environment.

*20% off for group bookings when booking 3 or more attendees from the same organization on the same course dates in the same transaction.

About this course

The AI Security course is designed to equip IT, cybersecurity, software, cloud, and other technical professionals with the knowledge and practical skills required to secure AI tools,  applications, and workflows in real-world business and technical environments. This hands-on program goes beyond basic awareness of AI risks and focuses on practical security assessment, access and data controls, and integrating AI security with established practices for protecting information, reviewing software, and managing operational risk.

Participants will learn how to map AI applications, identify trust boundaries, and assess risks across workflows involving AI assistants, organizational documents, external services, and tool-connected agents. Through realistic scenarios, guided analysis, and hands-on projects, learners will examine how information enters and leaves an AI workflow and evaluate risks such as misleading instructions, sensitive-information disclosure, and excessive permissions. Selected OWASP guidance for generative AI applications provides a reference for this work, with activities emphasizing defensive analysis in provided or explicitly authorized environments.

The course emphasizes safe, controlled, and accountable use of AI, helping teams translate security findings into practical safeguards, testing activities, and monitoring plans. By the end of the program, learners will be able to develop an AI Security Review and Control Plan that documents a system, identifies threats, recommends safeguards, and defines the conditions required before deployment or broader use. This final project brings the course concepts together in a provided or appropriately scoped workplace scenario, preparing participants to apply a structured approach to AI security within their organizations.

What you will learn

• Distinguish AI model behavior, application security, infrastructure security, and organizational responsibilities.
• Map the components, data flows, integrations, and trust boundaries of an AI-enabled application.
• Identify risks involving prompt injection, sensitive-information disclosure, untrusted content, and inappropriate tool use.
• Evaluate identity, permissions, service accounts, and access to organizational information.
• Define safeguards for data handling, secrets, retrieved content, AI outputs, and connected tools.
• Assess risks introduced by third-party models, software dependencies, and external services.
• Design defensive tests with expected behavior, evidence requirements, and clear limits.
• Develop logging, monitoring, escalation, and incident-response procedures for an AI workflow.
• Produce a security review that prioritizes findings, assigns ownership, and documents residual risk.

Topics Covered

Module 1: AI Security Foundations and Shared Responsibilities

• Distinguishing models, AI applications, infrastructure, and organizational processes.
• Comparing public AI services with internally integrated applications.
• Recognizing limitations of AI outputs and security claims.
• Connecting AI security with confidentiality, integrity, and availability.
• Identifying responsibilities shared across users, developers, operators, and providers.
• Defining authorization and safe boundaries for defensive activities.
• Hands-on exercise: Activity: Create an AI security context brief identifying the system, assets, responsibilities, assumptions, and authorized review scope.

Module 2: Mapping AI Applications and Trust Boundaries

• Identifying users, models, application services, data stores, and external integrations.
• Tracing inputs, retrieved information, outputs, and tool actions.
• Mapping identity and authentication points.
• Distinguishing trusted instructions from untrusted content.
• Identifying boundaries between organizational systems and external services.
• Documenting architecture assumptions and missing information.
• Hands-on exercise: Activity: Produce an architecture and data-flow package with components, assets, integrations, trust boundaries, and an evidence-gap log.

Module 3: Threat Modeling and Risk Prioritization

• Connecting assets and workflows to potential threat scenarios.
• Distinguishing vulnerabilities, threats, impacts, and safeguards.
• Examining misuse, unauthorized access, and unintended actions.
• Considering attacker-controlled inputs and external dependencies.
• Prioritizing findings using explicit likelihood and impact assumptions.
• Documenting control ownership and residual risk.
• Hands-on exercise: Activity: Create a threat model and prioritized risk register with evidence needs, affected assets, proposed controls, and owners.

Module 4: Protecting Data, Privacy, and Secrets

• Identifying sensitive inputs, retrieved information, generated outputs, and logs.
• Defining permitted data use and access boundaries.
• Examining retention, deletion, external processing, and disclosure questions.
• Keeping credentials and secrets outside prompts and inappropriate records.
• Applying data minimization and safe handling to test materials.
• Distinguishing security review from specialist privacy or legal review.
• Hands-on exercise: Activity: Develop a data-protection plan with handling rules, access boundaries, retention questions, and escalation needs.

Module 5: Prompt Injection and Untrusted Content

• Recognizing attempts to redirect application behavior through malicious instructions.
• Distinguishing direct user input from instructions embedded in retrieved or external material.
• Examining conflicts between intended tasks and untrusted content.
• Assessing effects on information disclosure and tool-connected actions.
• Designing layered controls rather than relying on a single prompt or filter.
• Defining defensive tests and documenting limitations.
• Hands-on exercise: Activity: Create a prompt-injection review with threat scenarios, defensive test cases, recommended controls, and remaining limitations.

Module 6: Identity, Access, and Least Privilege

• Reviewing user identities, service accounts, and application permissions.
• Separating authentication from authorization.
• Limiting access to data and connected systems according to task needs.
• Preserving user-specific permissions in retrieval and tool use.
• Defining credential-management and access-review expectations.
• Planning approval and revocation procedures.
• Hands-on exercises: Activity: Produce an identity and access-control plan with a permissions matrix, narrower access recommendations, and verification cases.

Module 7: Securing Retrieval and AI-Generated Outputs

• Mapping retrieval sources and document-access boundaries.
• Evaluating provenance, source relevance, and unauthorized information retrieval.
• Treating retrieved content and generated outputs as inputs requiring appropriate validation.
• Examining downstream uses in reports, applications, and automated processes.
• Defining output checks, safe handling, and human-review criteria.
• Documenting evidence and exceptions rather than assuming citations prove correctness.
• Hands-on exercises: Activity: Create a retrieval and output-control plan with source boundaries, validation requirements, review checkpoints, and test evidence.

Module 8: Securing Agents, Tools, and Automated Actions

• Mapping what an agent can read, change, send, or execute.
• Limiting available tools and action permissions.
• Separating proposed actions from approved execution.
• Defining human approvals for sensitive or consequential actions.
• Establishing transaction limits, safe failure behavior, and fallback procedures.
• Documenting action logs and operational ownership.
• Hands-on exercises: Activity: Develop an agent-control specification with allowed actions, permission boundaries, approval points, logging, and fallback procedures.

Module 9: Third-Party Components and Secure Delivery

• Identifying models, libraries, connectors, services, and other dependencies.
• Evaluating available vendor and component security evidence.
• Reviewing configuration changes, updates, and dependency risks.
• Applying code review, testing, scanning, and release controls where relevant.
• Protecting configuration, credentials, and deployment artifacts.
• Defining change-review and rollback expectations.
• Hands-on exercises: Activity: Produce a component and delivery review with dependency records, evidence requirements, release gates, and change-control procedures.

Module 10: Security Testing, Monitoring, and Incident Response

• Defining defensive test objectives, scope, and expected behavior.
• Selecting routine, edge-case, and failure scenarios.
• Collecting useful evidence while minimizing sensitive logging.
• Identifying monitoring signals and alert ownership.
• Planning containment, investigation, communication, and recovery.
• Retesting changes and documenting unresolved findings.
• Hands-on exercise: Create an assurance and response package with test cases, evidence requirements, monitoring criteria, and an incident-response procedure.

Module 11: AI Security Review and Control Plan

• Final exercise: For the final project, participants will work with an AI application scenario and complete a defensive security review. The scenario may involve a simulated document assistant, a tool-connected agent, or an organizational AI workflow. The project does not require production deployment, paid infrastructure, or testing of public services.
• Key Takeaways and next steps for attendees