Trust boundary review
Read the workflow, find untrusted inputs, and map the systems the agent touches.
Five-day AI agent security bootcamp
Secure the agents that read, decide, and act.
We train teams to scope deployed AI workflows, test agent failure modes, build controls, and prove readiness to a client security review board.
Bootcamp case arc
5 days
Scope the workflow
Map trust boundaries, authority, and blast radius.
Attack the agent
Test injection, tool misuse, and exfiltration paths.
Build controls
Use access maps, gates, guardrails, and evals.
Prove readiness
Assemble evidence for audit and review.
Course description
A practical security course for deployed AI agents.
Companies now place AI agents inside real workflows, where they read records, draft replies, and write to other systems. This course teaches AI FDSE skills: model the threats, scope access, add controls, watch production behavior, and give reviewers the evidence they need.
Adversarial testing
Run direct and indirect injection, privilege escalation, and exfiltration tests.
Control design
Build least privilege maps, approval gates, output checks, and sandbox boundaries.
Evidence and response
Use logs, evals, incident response, and compliance mapping to prove readiness.
Course schedule
One worked case carries through all five days.
Learner outcomes
Participants leave with review-ready evidence.
Boundary diagram
A trust boundary map for the agent, its inputs, tools, data, and authority.
Access map
A least privilege service-account map with allowed fields, denied fields, and proof of denial.
Security eval suite
A set of attack cases that runs beside accuracy evals and catches regressions.
Evidence package
A control package with logs, findings, eval results, and review-board talking points.
Who it is for
Built for technical and security teams near the deployment.
Target audience
Prerequisites
Best fit for learners with working AI knowledge, Python environment comfort, cybersecurity fundamentals, and enterprise context.
Preparation
Bring one real workflow from your job that you would like to secure. Expect hands-on work between sessions as each deliverable builds on the last.
Role signal
Current job listings show what the market expects.
These public listings show the skill mix behind FDSE work: security engineering, customer deployment, AI application delivery, controls, evaluation, and production ownership.
OpenAI
View listing
Forward Deployed Security Engineer
Security role focused on strategic public sector deployments, hands-on deployment security, infrastructure controls, monitoring, and validation.
Anthropic
View listing
Forward Deployed Engineer
Applied AI role focused on shipping Claude-powered applications, customer workflows, MCP servers, agent skills, deployment patterns, and production support.
Bring FDSE training to your team