About the job
About Our Team
At OpenAI, our Safety Systems organization is dedicated to ensuring the responsible development and deployment of our most advanced AI models. We create robust evaluations, safeguards, and safety frameworks that guarantee our models operate as intended in real-world environments.
The Preparedness team, an integral part of the Safety Systems organization, is guided by OpenAI’s Preparedness Framework.
While frontier AI models hold the promise of benefiting humanity, they also introduce significant risks. The Preparedness team is tasked with monitoring and preparing for catastrophic risks associated with these advanced AI models to ensure they drive positive change.
Our mission includes:
Proactively monitoring and assessing the evolving capabilities of frontier AI systems, focusing on identifying catastrophic risks.
Establishing concrete procedures, infrastructure, and partnerships to effectively mitigate these risks and manage the development of powerful AI systems safely.
This dynamic and impactful role connects capability assessments, evaluations, internal red teaming, and mitigations for frontier models, playing a crucial role in our overall AGI preparedness efforts.
About the Position
In this pivotal role, you will spearhead the Automated Red Teaming (ART) initiative, developing scalable, research-driven systems that continuously identify failure modes in our AI models and implement actionable improvements. Your primary goal will be to minimize potential harm by identifying the most critical vulnerabilities early and reliably.
Your Responsibilities
You will direct the research and technical strategy for automated red teaming across critical risk areas, focusing initially on:
Automated discovery of classifier jailbreak vulnerabilities (cybersecurity and biosecurity).
Automated elicitation of bio threat-development scenarios (worst-case planning uplift).