Jailbreak Pattern Catalogue

About this task

This Series B clinical data company for elder care (51–200 employees) is looking for a manager Red-Teamer for a rating and review programme for AI output in your own field, 3–8 h a week. Groups attacks into patterns that guide defences.

Ratings are checked against a written rubric, and your notes go back to the team that builds the product. It works in healthcare, so patient data is never part of the task and privacy rules are strict.

The work is fully remote. Any location in United States works, with a few hours of overlap with the team's day.

What you deliver

• Organise 120 successful attacks into a catalogue of patterns with examples.
• A weekly count of items reviewed, confirmed by the programme lead
• Flags for answers that are unsafe, off-topic or wrong in your field
• Ratings entered in the client's tool against the written rubric, with a one-line reason each

What a week looks like

• Tuesday: 15-minute check-in to agree the week's items
• Wednesday: main working block (3 hours)
• Monday: hand-off note with what shipped and what is open

Duration: Weekly batches for 6 weeks, renewed every four weeks. Accepted when the company contact confirms the deliverable against the written checklist. One round of edits is included.

Tools

How you get paid

• The company prepays Deftridge before your first hour, so payment does not depend on a late invoice.
• Deftridge pays you in USD after the company confirms the hours or the deliverable. The payment date is written in your task agreement.
• You work as an independent contractor: you report your own income, and your recruiter sends the statements your accountant needs.
• You never chase the company. Questions about money go to Deftridge.

What happens after you say you're interested

1. Within one working dayA recruiter writes to you to confirm your hours and the rate you want.
2. You agree the rateYou and the recruiter fix an exact rate inside the company's range. You decide whether to go ahead.
3. With your written OKThe company sees the anonymous profile below. Most companies answer within three working days.
4. StartYou sign the task agreement and the company's confidentiality terms through Deftridge, and begin on the date above.

What the company sees about you

First name, field, experience, tools and the hours you offer. Not your surname, your employer or your contact details.

How this rate compares

The midpoint of this range, $116 an hour, is above the median manager Red-Teamer rate in the United States ($110). The shaded band is the 25th to 75th percentile of the market.

All Red-Teamer jobs in United States · Red-Teamer rates in United States · AI Expert Review jobs

Questions about pay and privacy

The rules are the same for every task. Pricing explains the 15% fee, prepayment and when you are paid; How it works covers matching, what the company sees and side work next to a job.