Remote Adversarial AI Risk Evaluator (Red‑Team)

AuraOne; United States; On-site; USD 82,656 - 123,984; Full time

Job summary

AuraOne is seeking a Scheming Behavior Risk Evaluator for a remote contractor role to stress-test AI systems with adversarial prompts. Reviewers design attack scenarios, document failures with detailed reproduction steps, and tie each successful jailbreak to the specific rubric clause it violated so the safety team can patch the gap and regress-test the fix.

You will work asynchronously, contributing to a growing library of red-team rubrics, and coordinate with the safety team to triage new

Qualifications

• Experience red-teaming AI systems or adversarial ML work.
• Excellent written communication—your reports become the patch ticket.
• Able to document attempts and why in policy-grey areas.
• Familiar with prompt-injection, jailbreak, and policy-bypass taxonomies.
• Reliable async availability for at least 10 hours per week.

Responsibilities

• Design adversarial prompts to probe jailbreak, policy bypass, and prompt injection weaknesses.
• Document successful attacks with reproduction steps and violated policy clause.
• Score model defenses across single-turn and multi-turn conversations.
• Triage new attack vectors and route to safety team with severity ratings.
• Maintain a library of attack patterns and propose new red-team rubrics.
• Calibrate against the red-team cohort to keep coverage and severity consistent.

Skills

Adversarial prompting

Red-team analysis

Policy taxonomy

Failure documentation

Scheming Behavior Risk

AI safety

Red-team evaluation

Policy rubrics

Scheming

Behavior

Job description

AuraOne is seeking a Scheming Behavior Risk Evaluator for a remote contractor role to stress-test AI systems with adversarial prompts. Reviewers design attack scenarios, document failures with detailed reproduction steps, and tie each successful jailbreak to the specific rubric clause it violated so the safety team can patch the gap and regress-test the fix.

You will work asynchronously, contributing to a growing library of red-team rubrics, and coordinate with the safety team to triage new