Role Overview
Intelligence experts will simulate and evaluate how malicious actors could use large language models to plan, enhance, or operationalize acts of violence, terrorism, targeted assassinations, special operations, and tactical-level attacks. The role focuses on red teaming LLMs, assessing jailbreak attempts, and judging whether model outputs represent credible, material threats in real-world contexts.
Key Responsibilities- Conduct red teaming exercises against large language models to identify misuse scenarios and failure modes.
- Simulate how actors with malicious intent might leverage LLM outputs to plan or improve violent, terrorist, or targeted operations.
- Evaluate and interpret outputs from jailbreaking or adversarial prompts, assessing operational feasibility and threat credibility.
- Determine whether model-generated content constitutes a material risk in realistic operational contexts, and communicate those assessments clearly.
- Bachelor''s, Master''s, or PhD degree, or equivalent professional experience.
- Substantial professional background in intelligence, defense, cyber operations, counterintelligence, counterterrorism, or closely related fields.
- Experience with operational planning, threat assessment, or adversary emulation is highly relevant.
- Comfort assessing scenarios involving violence, targeting, and sensitive operational details.
- Location: Remote.
- Employment type: Part-time.
- Assignments are project-based and may involve irregular schedules tied to specific evaluation tasks.
Up to $200/hr (depending on the project)
Eligibility- Professionals with government or military experience in the listed fields are especially well suited to this role.
- Candidates must be willing to engage with content and scenarios that involve violent tactics and operational details for the purpose of risk assessment and mitigation.
This role is intended for experienced government, military, or operational professionals who want to apply their subject-matter expertise to anticipate national security risks posed by AI. Applicants will be evaluated on relevant operational experience and domain knowledge.