AI Red Team Tester

The work

Help improve AI safety by probing model defenses and documenting weaknesses. Design realistic adversarial tests, assess unexpected behavior and recommend practical improvements alongside engineering teams.

• Plan and carry out red team exercises targeting security weaknesses.
• Create adversarial prompts and unusual scenarios that test guardrails and safety filters.
• Assess outputs for unsafe behavior, bias and policy breaches.
• Write structured reports explaining vulnerabilities, exploits and unexpected results.
• Work with engineers on practical mitigations and improvements.
• Follow developments in AI threats, jailbreak methods and adversarial research.
• Help improve testing procedures and security assessment rubrics.

Experience and skills

• Strong cybersecurity knowledge in areas such as threat modeling, penetration testing or ethical hacking.
• Practical experience with AI or ML systems, language models or prompt engineering.
• Creative analytical judgment and an ability to recognize overlooked edge cases.
• Excellent technical writing and documentation.
• Comfort working independently on asynchronous assignments.
• Open source AI or LLM ecosystem familiarity is helpful.
• An information security, AI safety or adversarial ML background is an advantage, but not mandatory.

Working arrangements

The fully remote hourly contract involves 10 to 40 hours a week on a flexible schedule. The work offers opportunities to deepen AI security expertise, with continuing contracts and broader responsibilities potentially available.