Alice is seeking a Senior Researcher to drive research execution and technical delivery across frontier AI risks, with a focus on CBRNE and cybersecurity threat vectors.
AI tester with German language
Serbia, Slovakia, Romania +1
4 regions
May 14, 2024
Reposted Oct 31
Job description
Project Objective
This project aims to test and evaluate one AI agent in the development stage. By asking questions and requests, based on various categories (the training is provided), you will try to force the AI agent to say something harmful, offensive, dangerous or toxic an evaluate it’s responses.
The end goal is to prevent AI agent from sharing any harmful, offensive, dangerous or toxic content to the end users and make it safer.
This role is fully remote, full-time with a duration of several months (depending on the project progress it might be prolonged).
Full online training before the project will be provided.
Responsibilities:
- generate requests to the AI agent on the proposed topic;
- evaluate the response proposed by agent, classify it accordingly;
- meet daily targets;
- сommunication with a supervisor.
Requirements:
- fluent German;
- intermediate English (for training and internal communication with the team);
- basic tech savviness;
- good time-management skills;
- high level of responsibility;
- quick learner.
Available regions
Each region below is open to applicants in that country only. Apply through the link for your region.
Available in 4 regions: pick your country, then finish on the employer's site.
Truelancer AI is building a specialist red team for an AI search safety pilot with a global enterprise client. You’ll work as an individual contributor.
A prospective project for experts in chemistry, physics, nuclear science and energetic materials. Reviewers classify prompts as benign, dual-use or malicious, judge and rank model answers for safety, and write short rationales. Senior reviewers audit that work.
AfterQuery is looking for radioactive materials and health physics experts to help make AI models safer in their domain. You'll author and review evaluation tasks that test whether a model can tell a safe request from an unsafe one, and write the expert analysis behind each…