Read the pattern. Recognize the flaw.
Loading the latest lessons and your progress…
Loading the latest lessons and your progress…
Build the judgment of an AI Security Engineer.
Review the code that connects models, retrieval, tools, users, and sensitive data.
Explore the planned modules and their challenges below. Enrollment opens when the lectures and challenges are ready.
Explore syllabusWork through complete lessons and code-review challenges. Browse the curriculum and read lesson previews for free.
First week £1Cancel any time
Identify security failures in LLM applications, agents, tools, and model supply chains.
Each stage has evidence of practice. Reading an article on its own does not complete a code review step.
Complete the ordered lessons, guided exercises, and code review cases. Your Core progress counts completed steps in this syllabus.
42 steps across 8 modulesAfter the Core, solve fresh review cases when they come due. Mastery follows completion of the scheduled recall bank.
8 scheduled reviewsThe modules are designed to be completed in order. You can explore linked lessons and public practice cases now; path enrollment opens when the full sequence is ready.
Treat instructions from users and external content as untrusted data.
Preserve trust boundaries when retrieved content enters model context.
Constrain what model-selected tools may do.
Review MCP tool capabilities, inputs, and server trust.
Bind agent actions to the user, purpose, and least authority.
Validate model output before rendering or executing it.
Treat model artifacts and loaders as executable inputs.
Review the complete path from user input to model, retrieval, tools, and output.
Apply the Core skills to fresh cases without a vulnerability label.
Fresh cases are scheduled after the original practice. Completing these reviews earns mastery of the path.