AI security courses
Follow a fictional service from its first boundary map to a defensible review decision. Draft the artifacts, try the offline control exercises and compare your reasoning with worked feedback. Each course has free lessons you can read without an account or card.
Worked answers that start from a threat model, explain each control by mechanism, and include the follow-ups and scoring an interviewer uses.
START HERE, FREEMap the assistant’s data, authority and recipients20 min · no account needed→Choose by the work you want to do
Start with foundations if service identity, process isolation or credentials are unfamiliar. Otherwise choose the application you want to review. Work through that course's modules in order, adding to the same review packet.
Operate Application Security From Code to Deployment
You need to connect application findings to the delivery authority and deployed software they affect.
You build: An application inventory, executable evidence checks, an independent AI-patch review and a verified repair record.
Review Cloud Exposure, Identity and Data Security
You need to distinguish cloud posture, permissions, data sensitivity and actual activity in one review.
You build: A tested coverage register, versioned data-impact note, runtime interval calculation and staged integration decision.
Build Security Operations From Events to Response
You want to investigate from evidence and automate response without losing scope or uncertainty.
You build: Executable event and intelligence checks, a historical-search assessment, a response state machine and a worked recovery memo.
Security Fundamentals, From Packets to Platforms
You want to understand what connections, processes, identities and credentials actually protect.
You build: A service review with a boundary map, evidence gaps and acceptance tests.
Securing an LLM Feature, End to End
You can follow a web request and want to review an assistant that retrieves information and uses tools.
You build: A feature review from permission decisions through evaluation, release and incident response.
Secure an Engineering Agent, From Task to Release
You understand application access checks and want to reason about delegated, persistent agent work.
You build: An engineering-agent review packet with bounded work, effect accounting and a release decision.
Evaluate AI Security, From Protocol to Decision
You can follow an application boundary and want to judge whether an AI security result supports the decision being proposed.
You build: An assessment packet with a test contract, observation audit, weighted review, comparison and a revised recommendation.