Founding Security Researcher (AI Agent Security)

An early-stage AI security startup whose product stops dangerous AI agent actions inside large enterprises is looking for its first Security Researcher to define and lead its threat research program.

Responsibilities

• Probe popular AI coding agents and model harnesses the way a determined adversary would, from prompt injection to tool abuse and data exfiltration
• Run benchmarks comparing open and closed models against prompt injection and other attack techniques
• Publish research that security leaders read, remember and act on
• Turn findings into detections and enforcement policies that ship in the product, and investigate real agent incidents in customer environments
• Stay on top of new agent frameworks and emerging risks, and brief CISOs on what you find

Qualifications

• 2 to 10 years in security research, threat research, red teaming, detection engineering or AI security at a security company, enterprise security team or AI lab
• Hands-on AI attack work such as prompt injection, jailbreaks or tool abuse against models and agents
• A focus on threat research or red teaming rather than vulnerability management or AppSec alone
• Bonus: published research, CVEs, conference talks or CTF results

On-site in San Francisco, five days a week.