Posted 1 mo agoFirst seen 1 mo ago
SafetyTech Client #1 | Adversarial Task Writer for AI Security RL Gyms
·Serbia, Armenia, Bulgaria·via Breezy
Posted by SD Solutions on 5 October 2026, 46 days ago. Still on their Breezy board when we checked 3 min ago. Half of SD Solutions's roles come down within 21 days.
remoteFull time
What the posting is about
The role involves designing adversarial prompt injection scenarios in YAML for AI security testing. Candidates must have expertise in prompt injection, technical writing, and familiarity with Docker and CLI tools. Experience in pentesting, appsec, or LLM security is strongly preferred.
Your experience, ready for this role
Answer a few questions about your real work. Review the additions and download your CV.
Read out of the posting
LevelNot stated
Experience askedNot stated
EmploymentFull time
LocationSerbia, Armenia, Bulgaria
Work arrangementRemote
Visa sponsorshipNot stated
SalaryNot extracted, check the posting
Posted2026-10-05
Found viabreezy, direct from their system
We saw it 9 hours after it went up.
The posting, as the company wrote it
Employment: Full-Time
On behalf of SafetyTech Client #1 , SD Solutions is looking for a talented Adversarial Task Writer for AI Security RL Gyms.
SD Solutions is a staffing company operating globally. Contact us to get more details about the benefits we offer.
Responsibilities:
Design adversarial prompt injection scenarios in YAML for Alice’s RL Gym platform.
Run scenarios against frontier models, validate success rates, and refine tasks based on testing results.
Submit high-quality, validated tasks for review and acceptance.
Work on a flexible, freelance basis , choosing task contributions according to your availability.
Contribute to ongoing AI security evaluation projects, with opportunities to receive additional tasks based on your activity, consistency, and task quality.
Requirements:
Adversarial mindset: you think like an attacker and understand how to exploit an AI agent’s helpfulness, authority assumptions, or trust in its environment
Prompt injection expertise: direct (role-play, encoding, context flooding) and indirect/environment-embedded (poisoned tool responses, malicious content in documents, cross-context leakage)
Technical writing in YAML
Comfortable with Docker, CLI tools, and running systematic tests against multiple models
Domain realism in at least one vertical: e-commerce, finance, HR, enterprise SaaS, healthcare, travel
Background in pentesting, appsec, LLM security research, or red teaming strongly preferred
The Task
You build adversarial prompt injection tasks for Alice’s RL Gym platform. Each task is a self-contained YAML scenario simulating a realistic AI agent deployment, testing whether the agent can be manipulated into violating its safety policies.
This is a freelance, task-based opportunity with flexible workload and no fixed weekly commitment. You can contribute according to your availability, with the possibility of receiving more tasks over time based on your activity and the quality of your submissions.
About the company:
A company building specialized evaluation infrastructure for AI safety and robustness testing. Their platform simulates adversarial conditions used by AI development teams to validate agent behavior before deployment. Currently expanding a freelance contributor pool for scenario and environment development.
By applying for this position, you agree to the terms outlined in our Privacy Policy. Please take a moment to review our Privacy Policy https://sd-solutions.breezy.hr/privacy-notice , and make sure you understand its contents. If you have any questions or concerns regarding our Privacy Policy, please feel free to contact us.
Copied from SD Solutions’s own board, not rewritten.
Also open at SD Solutions
Why this page exists
We read companies’ own hiring systems every hour, 2,106 of them, and show a job the hour it opens instead of when a job board gets around to indexing it. We saw it 9 hours after it went up.
The feed is free. No card, no trial to expire.
A free account and your résumé first, no card