LLM Security / AI Red Team Specialist – Freelance, United States

We are seeking experienced LLM Security / AI Red Team Specialists based in the United States to support AI safety and security projects focused on evaluating the robustness of large language models and AI systems.

Responsibilities

• Design and evaluate adversarial test cases for large language models and AI systems.
• Identify weaknesses in model behavior, including prompt injection, indirect prompt injection, jailbreaks, instruction hierarchy failures, and other adversarial patterns.
• Create realistic scenarios that reflect how AI systems interact with external content such as webpages, documents, emails, search results, structured data, and other digital artifacts.
• Assess whether model responses create security, safety, policy, or misuse risks.
• Classify and annotate test cases according to detailed project guidelines.
• Write clear rationales explaining why a scenario is technically and contextually plausible.
• Review and improve test cases for realism, diversity, quality, and duplication.
• Work with QA or project leads to resolve ambiguous or edge-case scenarios.
• Follow detailed security, safety, and data-handling requirements consistently.

Requirements

• Currently reside in the United States and have continuously resided in the United States for the past five consecutive years.
• Strong understanding of large language models, generative AI, and common LLM security risks.
• Familiarity with concepts such as:
• Prompt injection
• Indirect prompt injection
• Jailbreaking
• Instruction hierarchy and model behavior
• Data exfiltration or unauthorized disclosure risks
• Tool-use and agent-based AI risks
• Strong analytical and problem-solving skills.
• Excellent written English and ability to explain technical reasoning clearly and concisely.
• High attention to detail and ability to follow complex written guidelines.
• Ability to work independently and manage project-based workloads.
• Comfortable working with potentially sensitive or adversarial content.

Preferred Experience

Experience in one or more of the following areas is highly desirable:

• AI red teaming
• LLM evaluation
• Application security
• Cybersecurity
• Adversarial testing
• AI safety
• Prompt engineering
• Security research
• Trust & Safety
• Technical QA
• Data annotation for AI or machine learning

Practical experience with web content, JSON, APIs, emails, PDFs, search results, or other structured/unstructured digital content is also beneficial.

Qualification Process

Applicants may be required to complete a short qualification assessment designed to evaluate their understanding of LLM security concepts, ability to identify realistic adversarial scenarios, and ability to follow project-specific guidelines.

Successful completion of the qualification process may be required before access to live project work is provided.

This is a flexible freelance opportunity, typically requiring 5–10 hours per week. Hours may vary based on project needs and task availability, and assignments may range from short-term projects to longer-term engagements.

Benefits:

• Flexible schedule

Work Location: Remote