AI Security Tester

Job summary

Conduct red-team tests of AI systems using adversarial prompts, jailbreak attempts, and edge cases to identify vulnerabilities, unsafe outputs, and policy failures. Document findings, recommend mitigations to engineering and research teams, and help develop security evaluation frameworks and testing protocols.

Job details

About The Role AI systems are only as safe as the people who challenge them. We're looking for security-minded professionals to stress-test, probe, and harden cutting-edge AI models. If you enjoy finding weaknesses, crafting clever exploits, and thinking like an attacker — this role was built for you. Your work will have a direct impact on the safety and reliability of AI products used by millions of people worldwide. Organization: Alignerr Type: Hourly Contract Location: Remote Commitment: 10–40 hours/week What You'll Do Conduct red-teaming exercises to uncover security weaknesses and failure modes in AI systems Design and execute adversarial prompts, jailbreak attempts, and edge-case scenarios to test model guardrails Evaluate AI outputs for safety, bias, harmful content, and policy compliance Document vulnerabilities, unexpected behaviors, and exploits in clear, structured reports Collaborate with engineering and research teams to recommend practical mitigations Stay sharp on emerging AI security threats, attack techniques, and evolving best practices Help define and refine security evaluation frameworks and testing protocols Who You Are Solid understanding of cybersecurity concepts — threat modeling, attack vectors, or penetration testing Hands-on experience with AI/ML systems, large language models (LLMs), or prompt engineering Creative and analytical — you love breaking things methodically to make them more robust Strong written communication and documentation skills Self-directed and comfortable working asynchronously on task-based assignments Background in infosec, ethical hacking, or AI safety research is a strong plus Familiarity with open-source AI platforms or red-teaming tools is a bonus Why Join Us Work at the forefront of AI safety alongside top research labs and engineering teams Directly shape how AI systems protect users — at scale, in production Fully remote with complete schedule flexibility — work when and how you want Build deep, in-demand expertise in one of the fastest-growing domains in tech Ongoing contract potential with opportunities for expanded scope and responsibility Join a global community of experts making AI more trustworthy for everyone

What you’ll do

Conduct red-team tests of AI systems using adversarial prompts, jailbreak attempts, and edge cases to identify vulnerabilities, unsafe outputs, and policy failures. Document findings, recommend mitigations to engineering and research teams, and help develop security evaluation frameworks and testing protocols.

Requirements

Candidates should understand cybersecurity concepts such as threat modeling, attack vectors, or penetration testing, and have hands-on experience with AI/ML systems, large language models, or prompt engineering. Strong analytical and written communication skills, self-direction, and comfort with asynchronous task-based work are expected; infosec, ethical hacking, or AI safety research experience is a plus.

Benefits

• Fully Remote Work
• Flexible Schedule
• Ongoing Contract Potential
• Opportunities For Expanded Scope And Responsibility
• Professional Development

Listed skills

• Technical Documentation · Preferred
• prompt engineering · Preferred

Other relevant skills

Identified from the job description. Confirm important requirements above.

• Cybersecurity
• Threat Modeling
• Penetration Testing
• AI/ML Systems
• Large Language Models
• Prompt Engineering
• AI Red Teaming
• Adversarial Prompting
• Jailbreak Testing
• Vulnerability Assessment
• AI Safety Evaluation
• Safety And Policy Compliance
• Technical Documentation
• Analytical Thinking
• Asynchronous Collaboration

Job areas

• Security & Safety
• Technology
• Software
• Science & Research

More jobs from Alignerr