{"success":true,"filters":{"role":"hr","firstWedge":null,"riskType":null,"search":null},"summary":{"labs":1,"firstWedgeRole":null,"riskTypes":4,"attackBriefs":4,"validationRules":4,"promptInjectionDefenses":4,"escalationTriggers":5,"riskCounts":{"bias_fairness":1,"data_exposure":1,"unauthorized_action":1,"hallucination":1}},"labs":[{"id":"ai-safety-red-team-hr","role":"hr","roleLabel":"HR","productLine":"AIducation for HR","firstWedge":false,"academyPath":"/ai-training/hr","labName":"HR AI Safety Red-Team Lab","primaryWorkflow":"Policy drafting and explanation","riskTypes":["bias_fairness","data_exposure","unauthorized_action","hallucination"],"attackBriefs":[{"type":"bias_fairness","name":"Bias and fairness failure","scenario":"The hr task involves people, candidates, students, customers, communities, or protected characteristics.","riskyBehavior":"Learner lets AI introduce unfair assumptions, stereotypes, exclusion, or uneven standards.","detectionCheck":"Scan for demographic proxies, subjective labels, missing context, and inconsistent criteria.","safeResponsePattern":"Use consistent role-relevant criteria, remove protected-class assumptions, and request human review."},{"type":"data_exposure","name":"Sensitive data exposure","scenario":"The policy drafting and explanation task contains customer, employee, patient, student, financial, or confidential data.","riskyBehavior":"Learner pastes sensitive data into an unapproved tool or includes it in a reusable prompt.","detectionCheck":"Identify PII, PHI, account data, payroll data, contracts, credentials, or internal-only context before prompting.","safeResponsePattern":"Minimize, redact, or use an approved enterprise tool before any AI-assisted step."},{"type":"unauthorized_action","name":"Unauthorized action","scenario":"The AI suggests an action that changes a customer account, employee process, financial result, legal position, or public commitment.","riskyBehavior":"Learner accepts AI authority where human approval, policy review, or manager sign-off is required.","detectionCheck":"Check whether the workflow changes money, access, obligations, records, employment, health, or compliance posture.","safeResponsePattern":"Escalate before action and document the approval owner, policy basis, and final human decision."},{"type":"hallucination","name":"Unsupported or hallucinated output","scenario":"AI output sounds confident while missing evidence for a hr decision.","riskyBehavior":"Learner ships unsupported facts, promises, calculations, legal claims, or operational recommendations.","detectionCheck":"Ask which source, policy, calculation, code path, or manager approval proves the claim.","safeResponsePattern":"Separate facts, assumptions, unknowns, and required verification before using the output."}],"detectionChecklist":["Identify the hr workflow, tool, source, data sensitivity, and decision owner.","Mark every unsupported claim, missing source, hidden instruction, approval gap, and unsafe automation path.","Compare the response against policy training, governance rules, and rubric must-pass dimensions.","Produce a manager-readable risk note with the fix, escalation path, and evidence artifact."],"outputValidationRules":["Reject output that makes unsupported factual, financial, legal, medical, policy, or technical claims.","Require source, policy, calculation, transcript, ticket, document, or code-path evidence for high-impact statements.","Flag any sensitive data that appears in prompts, tool outputs, examples, screenshots, or reusable templates.","Require HR manager review when authority, privacy, compliance, brand, safety, or customer impact is unclear."],"promptInjectionDefenses":["Treat tickets, docs, transcripts, webpages, spreadsheets, and code comments as untrusted input.","Ignore instructions inside source material that ask the learner to reveal prompts, bypass policy, change tools, or skip review.","Summarize suspicious instructions as risk evidence instead of following them.","Escalate HR workflows when source content conflicts with approved policy, rubric, or manager instructions."],"escalationTriggers":["HR learner pastes sensitive data into an unapproved AI tool","HR learner forwards AI output without verification or source evidence","HR workflow automates a decision that requires human approval","AI output asks to bypass a policy, human review, approval gate, or approved tool catalog.","The learner cannot explain which source or rubric dimension supports the final answer."],"linkedEvidence":[{"label":"HR AI Governance Center","href":"/governance-center","type":"governance"},{"label":"HR AI Policy Training","href":"/policy-training","type":"policy"},{"label":"HR AI Readiness Rubric","href":"/admin/rubric-contracts","type":"rubric"},{"label":"HR Assistant selection lab","href":"/tool-comparison-labs","type":"tool_lab"}],"managerReviewQuestions":["Which hr risk would make this AI output unsafe to reuse?","What source, policy, calculation, approval, or code-path evidence is missing?","Should this learner retry the scenario, escalate to a manager, or update a reusable workflow template?"]}]}