AI Jailbreak & Prompt-Injection Security Expert

The work

Evaluate AI safety through authorized adversarial testing, prompt injection research and tool abuse scenarios. This remote contract turns security expertise into structured assessments and practical improvements for model robustness.

• Develop evaluation methods for ethical jailbreak testing, LLM red teaming, prompt injection and misuse of connected tools.
• Create elicitation strategies across domains to investigate complex bypasses and attacks spanning multiple conversation turns.
• Build and maintain regression suites that check susceptibility to jailbreaks and injected instructions.
• Construct assessment frameworks reflecting realistic adversarial threats.
• Translate findings into concrete safety and risk reduction recommendations with technical stakeholders.
• Explain methods, results and effective practices in reports and presentations suited to technical and general audiences.

Preferred experience and skills

• Preferred experience is at least two years in adversarial ML, LLM red teaming, AI safety testing or closely related security work.
• A record of discovering, studying or testing prompt injection, ethical jailbreaks, tool misuse or other adversarial AI weaknesses.
• An MS or PhD in computer science, cybersecurity, ML or a relevant subject, or comparable professional or operational experience.
• Community credibility through research publications, open source tools or conference talks is valued.
• Strong documentation and spoken communication, with collaborative problem solving skills.
• Interdisciplinary safety project experience and familiarity with current LLM designs, prompt methods and security tools are advantages.

Working arrangements

This contractor role is remote. A prior AI employment background is not mandatory, although relevant adversarial security expertise is preferred.