0/35FDE path lessons understood. Go to your dashboard.

 

Topic 06 of 10

Security & risk review

What every security review asks: prompt injection, tool poisoning, threat models, data handling and the questionnaires that stand between you and go-live.

  • 10 lessons
  • 5.9 h reading
  • 11.4 h hands-on
  • 2 on the FDE path
Start lesson 01
Progress0 / 10

Lessons

  1. 01Guardrails the risk team can sign LabLayered safety checks, each with an action, a threshold and a log. Pro
  2. 02Defending against prompt injectionPath · stage 5Every document and tool result is untrusted input to an agent with tools. Pro
  3. 03Automated red-teamingAttack your own agent at scale before an auditor or a journalist does. Pro
  4. 04Bias and fairness evidenceSpecialisationMeasure and report fairness where hiring, lending and public services demand it. Pro
  5. 05Refusals and safe failureDecide what the agent does when it shouldn't act, and make that graceful. Pro
  6. 06SSO and provisioning for agentsGive the agent an enterprise identity before it gets a prompt. Pro
  7. 07Authorization and multi-tenancySpecialisationRBAC, ABAC and tenant isolation: the least privilege auditors actually check. Pro
  8. 08Agent credentials, from issue to revokeNon-human identities that don't outlive the project as standing risk. Pro
  9. 09Secrets, dependencies and network isolationNo hardcoded keys, pinned dependencies, and agents fenced off the network. Pro
  10. 10Threat modelling an AI systemPath · stage 5A shared vocabulary with the security team, and a register they can track. Pro