About the job
About Our Team
The Safety Systems organization at OpenAI is dedicated to ensuring the responsible development and deployment of our most advanced AI models. We create evaluations, safeguards, and safety frameworks to ensure that our models operate as intended in real-world scenarios.
The Preparedness team plays a critical role within the Safety Systems organization, guided by OpenAI’s Preparedness Framework.
As frontier AI models hold the potential for significant benefits to humanity, they also present escalating risks. The Preparedness team is essential in preparing for the development of increasingly capable frontier AI models, focusing on identifying, tracking, and preparing for catastrophic risks associated with these technologies.
The mission of the Preparedness team includes:
Monitoring and predicting the evolving capabilities of frontier AI systems, especially regarding risks that could have catastrophic consequences.
Establishing concrete procedures, infrastructure, and partnerships to mitigate these risks and safely advance the development of powerful AI systems.
The Preparedness team integrates capability assessment, evaluations, internal red teaming, and mitigations for frontier models, coordinating overall AGI preparedness. This fast-paced and impactful work holds significant importance for both our organization and society.
About the Role
As frontier AI systems become more capable, they exhibit greater autonomy, the ability to pursue long-term goals, adapt to feedback, and utilize tools. While these advancements offer immense potential, they also raise the risk of models behaving in misaligned or deceptive ways, which can be difficult to supervise or contain. Addressing the risk of loss of control is a key challenge in the safe development and deployment of advanced AI systems.
In your role as a Researcher focused on loss of control mitigations, you will design and implement an end-to-end mitigation strategy aimed at reducing the risk of intentionally subversive or inadequately controllable behaviors in OpenAI’s products and internal operations. This position necessitates strong technical expertise and close collaboration across functions to ensure that safeguards are enforceable, scalable, and effective. You will directly contribute to establishing robust protections as model capabilities evolve.