Note: The job is a remote job and is open to candidates in USA. Crossing Hurdles is seeking an AI Jailbreak & Prompt-Injection Security Expert to evaluate and improve the safety of AI systems. The role focuses on adversarial testing, vulnerability research, regression testing, evaluation framework development, and documenting security findings for stakeholders.
Responsibilities
- Design and implement advanced methodologies for evaluating AI system safety
- Create comprehensive cross-domain elicitation strategies for adversarial bypass patterns
- Develop and maintain regression test suites for jailbreak and prompt-injection vulnerabilities
- Construct evaluation frameworks to stress-test AI models against adversarial threats
- Collaborate with stakeholders to translate security findings into model safety improvements
- Document methodologies and findings in clear reports and presentations
Skills
- Have strong relevant experience in adversarial machine learning or AI safety evaluation
- Proven experience in researching vulnerabilities related to ethical jailbreaks or adversarial AI attacks
- Advanced degree in computer science, cybersecurity, or a relevant discipline, or equivalent experience
- High credibility within the AI security community, such as published research or open-source tools
- Exceptional written and verbal communication skills with a focus on documentation and problem-solving
Benefits
- Remote
Company Overview