Harmful Instruction Refusal Risk Evaluator

AI Trainer Jobs; United States; Remote; USD 90,000 - 96,000; Full time

Job summary

AI Trainer Jobs seeks a Harmful Instruction Refusal Risk Evaluator for a remote red-team track that stress-tests AI systems against adversarial prompts. Reviewers craft attack scenarios, document failures, and pair each jailbreak with the violated rubric clause to help patch gaps.

AuraOne safety team collaborates to harden models before shipping. Adversarial evaluation is performed with rigorous write-ups, enabling reproducibility and regression testing by the modeling team.

Qualifications

• Proven red-team adversarial testing experience.
• Document attack scenarios with reproducible steps.
• Score defenses across single-turn and multi-turn conversations.
• Available for at least 10 hours per week.

Responsibilities

• Design adversarial prompts that probe known weakness classes.
• Document every successful attack with reproduction steps.
• Score model defenses across conversations.
• Triage emerging attack vectors for safety teams.

Skills

Adversarial prompting

Red-team analysis

Policy taxonomy

Failure documentation

Harmful instruction refusal

AI safety

Red-team evaluation

Policy rubrics

Job description

Harmful Instruction Refusal Risk Evaluator is a remote red-team track for stress-testing AI systems against adversarial prompts. Reviewers craft attack scenarios, document the failure mode, and pair each successful jailbreak with the rubric clause it violated so the safety team can patch the gap.

Category: AI Safety & Red Teaming Pay: $65–$70 / hr Location: Remote — US-eligible Contractor

Harmful Instruction Refusal Risk Evaluator is a remote red-team track for stress-testing AI systems against adversarial prompts. Reviewers craft attack scenarios, document the failure mode, and pair each successful jailbreak with the rubric clause it violated so the safety team can patch the gap.

About the role

Harmful Instruction Refusal Risk Evaluator is a remote red-team track for stress-testing AI systems against adversarial prompts. Reviewers craft attack scenarios, document the failure mode, and pair each successful jailbreak with the rubric clause it violated so the safety team can patch the gap. Adversarial evaluation is how AuraOne hardens AI models before they ship to customers. Reviewers think like attackers and write up failures with enough rigor that the modeling team can reproduce, fix, and regress-test them. Push on refusal boundaries and dual-use risk before a model ships.

Responsibilities

• Design adversarial prompts that probe known weakness classes (jailbreak, policy bypass, prompt injection) for Harmful Instruction Refusal Risk Evaluator assignments.
• Document every successful attack with reproduction steps and the policy clause it violated.
• Score model defenses across single-turn and multi-turn conversations.
• Triage emerging attack vectors and route them to the safety team with severity ratings.

Role details

Track Adversarial evaluation Work model Remote Independent specialist contractor Compensation Hourly rate confirmed after the interview process. Eligible from US

What you should bring

• Demonstrated experience red-teaming AI systems, security research, or adversarial ML work for Harmful Instruction Refusal Risk Evaluator work.
• Strong written communication — your reports become the patch ticket.
• Comfort working in policy-grey areas with clear documentation of what was attempted and why.
• Familiarity with prompt-injection, jailbreak, and policy-bypass taxonomies.
• Reliable async availability for at least 10 hours per week.

Example tasks

• Construct a 5-turn adversarial conversation that bypasses a specific policy clause and write up the patch ticket.
• Score a model's defenses against a known jailbreak pattern across 20 variants.
• Propose a new red-team rubric category after spotting an emerging attack vector.
• Reproduce a failure another reviewer reported and confirm the severity tag.

Useful experience

• Background in offensive security, AppSec, or trust & safety operations.
• Experience publishing or reproducing public adversarial-ML research.
• Multilingual fluency for cross-language attack testing.

Compensation and schedule

Hourly rate confirmed after the interview process.
Expected arrangement: contractor , with program-defined task volume and review pacing. Placement depends on current program demand and reviewer confirmation.

Skills used in matching

• Adversarial prompting
• Red-team analysis
• Policy taxonomy
• Failure documentation
• Harmful Instruction Refusal Risk evaluation
• AI safety
• Red-team evaluation
• Policy rubrics
• Harmful
• Instruction
• Refusal