OpenAI Offers $445,000 Research Role Focused on Self-Improving AI Risks

OpenAI has posted a research position on its Preparedness safety team, offering compensation of up to $445,000, according to HRKatha. The role focuses on anticipating potential risks from advanced AI systems that could improve their own capabilities without direct human oversight.

The position, described as a Researcher in Recursive Self-Improvement Preparedness, involves studying speculative risks such as recursive self-improvement, data poisoning, and automation of technical work. Researchers will design experiments to monitor unexpected AI behavior, develop defenses against manipulated training data, and build tools to interpret model reasoning. The job also includes tracking how automation may begin to replace technical roles within the company.

As detailed in a posting on OpenAI's website, the Preparedness team focuses on measuring and mitigating risks from frontier AI systems and maintaining the company’s safety framework. Candidates are expected to execute technical work quickly, build prototypes, and coordinate with other teams to prepare for potential future security threats.

This position marks a continuation of OpenAI’s investment in safety research as it seeks to address emerging challenges linked to the rapid evolution of AI capabilities.