About the team

Ensuring that advanced AI benefits humanity requires protecting the research, infrastructure, and trust that make its development and deployment possible. OpenAI Security defends our model IP, people, products, and users against sophisticated adversaries across GPU supercomputing, multicloud environments, datacenters, and emerging agent systems. The Security Incident Response Team (SIRT) leads the response when those defenses are tested—establishing what happened, limiting harm, and strengthening our ability to protect frontier AI.

About the role

We’re looking for a hands-on technical leader to build the incident response team that protects OpenAI as AI capabilities and the threats surrounding them evolve. You will command consequential incidents, lead technically demanding investigations, and make containment and recovery decisions when evidence is incomplete and the stakes are high. You will also shape how frontier models change incident response: using OpenAI’s own technology to help responders connect evidence, test hypotheses, and act effectively at a scale that manual workflows cannot support. Partnering with Research, Infrastructure, Product Security, Enterprise Security, IT, and Engineering, you will build exceptional responders, dependable AI-powered capabilities, and stronger defenses through every investigation.

In this role, you will

Hire, coach, and develop exceptional incident responders who combine investigative depth, engineering ability, and sound judgment. Build a team that challenges assumptions, surfaces uncertainty early, and uses AI to expand its capabilities. Develop technical leaders who can make and explain difficult decisions under pressure.

Lead incident command for complex threats to OpenAI’s research, products, and infrastructure. Establish severity and scope, coordinate investigation workstreams, and weigh containment options against ongoing exposure and disruption to research or production.

Build and ship agentic capabilities using OpenAI’s models for evidence collection, investigation context, response coordination, and bounded response workflows. Partner with platform engineers to integrate tools and data, evaluate accuracy and failure modes, and ensure reliable, auditable operation with human approval for consequential actions.

Set investigation standards for sophisticated adversaries and insider threats across identity, endpoints, cloud, Kubernetes, and large-scale compute. Help responders reconstruct activity, test competing hypotheses, preserve evidence, and make uncertainty explicit.

Drive incidents to verified closure. Work with accountable owners to validate containment and recovery, document remaining risk, and track corrective actions that reduce the likelihood or impact of recurrence.

Coordinate with engineering, infrastructure, legal, privacy, and other stakeholders throughout an incident. Give executives clear updates on impact, uncertainty, decisions, and the next update without overstating what the evidence supports.

Build sustainable on-call coverage, escalation paths, and handoffs across regions. Balance urgent response with training and planned improvements, and protect the team’s ability to recover after demanding incidents.

Build readiness into new research and infrastructure capabilities with their owners. Exercise playbooks, escalation paths, and containment options, including scenarios involving agents operating across systems at scale.

Turn incident evidence into stronger defenses with Detection Engineering, Threat Intelligence, Security Observability, and security platform teams. Identify what should have exposed an intrusion earlier, what prevented a faster or more confident investigation, and which capabilities would have changed the outcome. Translate those findings into tested improvements with clear owners.

Prioritize improvements using incident outcomes and readiness evidence. Measure response quality and effectiveness, make tradeoffs explicit, and address recurring friction without reducing success to ticket volume or speed alone.