OpenAI has announced a new research position aimed at addressing potential risks from rapidly advancing artificial intelligence. The company is offering up to $445,000 annually for the role, which will focus on ensuring that AI systems do not become excessively self-improving beyond human oversight. The position falls under OpenAI’s Preparedness safety team, which studies long-term risks associated with increasingly powerful AI systems.

According to the job description, candidates must be both technically skilled and strategically thoughtful, as the work involves anticipating risks that may not yet exist but could emerge quickly. A key focus area is “recursive self-improvement,” where an AI could autonomously design and train more advanced versions of itself. The researcher will identify and mitigate such risks, conduct safety tests on advanced models, and develop monitoring systems to detect abnormal behavior.

The role also includes protecting AI models from “data poisoning” attacks and creating tools to better understand how advanced AI systems make decisions. The announcement comes amid growing industry discussion about recursive self-improvement, with leaders from Google DeepMind and other institutions warning of accelerating AI capabilities.