AI Red Teamer & Safety Evaluator (all genders)
Remote (EU time zones) / Remote (worldwide)
Freelance · Part-time · Full-time
Open until 30 Nov 2026
About the role
Break things so nobody else can. You systematically probe AI agents for jailbreaks, prompt injections, data leaks and harmful behaviour – and help us build governed agents that enterprises and public institutions can trust.
What you’ll do
- Design and run red-teaming campaigns against LLMs and tool-using agents.
- Test for prompt injection, data exfiltration, privilege escalation via tools and policy violations.
- Document findings with reproducible steps, severity and mitigation proposals.
- Build reusable adversarial test suites for continuous evaluation.
- Work with engineering on guardrails, policies and human-in-the-loop flows.
What you bring
- Background in security, trust & safety, AI research or a related field.
- Deep understanding of how LLMs, RAG and agent tool-use work.
- Creative, adversarial mindset – combined with strong ethics.
- Excellent written English; German is a plus.
- Ability to write clear, structured reports.
Nice to have
- Pentesting or bug bounty experience.
- Knowledge of the OWASP Top 10 for LLM applications and the EU AI Act.
What we offer
- Fully remote and asynchronous – work when and where you want.
- Fair, transparent hourly rates, paid reliably every month.
- Structured onboarding, clear guidelines and fast feedback from our quality leads.
- Varied projects: our own AI agents, client models and evaluation suites.
- A long-term collaboration – the best trainers grow into lead and reviewer roles.
- See first-hand how frontier AI is built, evaluated and made safe.
Tools & stack
ClaudeGPTMCPPythonPromptfooBurp Suite
Apply now
Apply for this role
AI Red Teamer & Safety Evaluator (all genders) · kalAIDoCreate
Takes about 3 minutes. CV or LinkedIn profile is enough.
