security complianceannotation qualitycontent policyllm red teamingtrust and safety
Complete evaluation framework
What to assess and how to score it
Review the evidence signals before interviewing. Then use the anchored descriptions—not instinct alone—to choose the score that best matches each answer.
01
Evaluation factor
Technical depth
35% weight
Check command of policy taxonomies and moderation tooling: labelling guidelines, classifier confidence thresholds, queue triage in platforms like Sift or internal review consoles, plus DSA and COPPA scope.
Evidence to listen for
Command of the specific attack surface, tooling, and controls the role covers
Understands how the underlying system works, not just how the tool reports on it
Can explain an attack or control chain end to end
Distinguishes what they found themselves from what a scanner flagged
Five-point scoring guide
1
Poor
Tool operator only; no understanding of the systems underneath.
2
Needs Improvement
Runs tooling but cannot explain findings or how the attack works.
3
Satisfactory
Solid working knowledge; depth thins outside familiar tooling.
4
Very Good
Strong command of the domain; explains attack and control chains clearly.
5
Excellent
Cites specific policy lines they enforced, explains how classifier scores routed items to human review, and names inter-rater agreement targets.
02
Evaluation factor
Real incidents and findings
30% weight
Probe actual review volume and hard cases: escalated self-harm or violent extremism queues, coordinated inauthentic behaviour waves, LLM jailbreak prompts they red-teamed and documented.
Evidence to listen for
Brings specific incidents, findings, or audits they personally worked
States their own role rather than the team's
Describes what was actually at risk and what changed afterwards
Can talk about a finding that turned out to be wrong
Five-point scoring guide
1
Poor
No hands-on work; knowledge is entirely certification or coursework.
2
Needs Improvement
Limited exposure; cannot describe their contribution to an incident.
3
Satisfactory
Real casework with adequate detail; ownership sometimes vague.
4
Very Good
Specific incidents with clear personal scope and what changed after.
5
Excellent
Describes named incidents with volumes, escalation paths, and the label or policy change that followed their write-up.
03
Evaluation factor
Risk judgement
20% weight
Assess handling of grey areas: satire versus harassment, medical misinformation, cultural context in non-English content, and where they chose leave-up with label over removal.
Evidence to listen for
Prioritises by actual exploitability and business impact, not raw severity scores
Can argue for accepting a risk as well as fixing it
Knows the difference between a finding and a problem
Does not cry wolf or wave things through
Five-point scoring guide
1
Poor
Treats every finding as critical, or waves real risk through.
2
Needs Improvement
Follows severity scores mechanically; no business context.
3
Satisfactory
Reasonable prioritisation; less confident arguing for risk acceptance.
4
Very Good
Prioritises by exploitability and impact; can justify accepting a risk.
5
Excellent
Reasons from harm severity and precedent rather than gut feel, and admits calls later overturned on appeal with the lesson taken.
04
Evaluation factor
Getting things fixed
15% weight
Look for evidence they closed loops: guideline rewrites, training data corrections fed to model teams, false-positive rate reductions, appeal backlog cleared, wellness and rotation practices.
Evidence to listen for
Writes findings engineers can act on rather than a wall of output
Has persuaded a team to fix something they did not want to fix
Explains risk to executives in business terms
Works with the org rather than policing it
Five-point scoring guide
1
Poor
Adversarial with engineering; findings never get fixed.
2
Needs Improvement
Reports are unactionable; no influence beyond raising tickets.
3
Satisfactory
Adequate reporting; relies on mandate rather than persuasion.
4
Very Good
Actionable findings and a real record of getting fixes shipped.
5
Excellent
Shows a guideline or dataset they changed, with measured drop in enforcement errors or appeal overturns after the fix.
Put this rubric to work
Score every candidate against the same standard
Add these weighted factors to Hirevire and let AI evaluate recorded answers against your rubric.
