AI Security & Control Researcher
Core
Design threat models and control protocols for coding agents to prevent misalignment and compromise, improving the security of the Watcher monitoring product.
Role type
Senior individual contributor security researcher (AI security)
Builds
Threat models, security levels, failure mode libraries, and adversarial test suites for AI agents
Domain
AI security, coding agents, threat modeling
Deliverable
production ML models
Required skills
threat modeling, offensive security, security research, code reading, infrastructure evaluation, adversarial thinking, technical controls design
Preferred skills
AI/ML systems security, LLM security, detection engineering, insider threat programs, red teaming
Technologies
MITRE ATT&CK, Watcher product
Responsibilities
Develop threat models for coding agents assuming misalignment or compromise; maintain a library of coding agent failure modes; design realistic attack trajectories for monitor backtesting; adjudicate flagged agent trajectories; red-team and improve Watcher monitors and policies
Seniority
Senior, hands-on IC