OpenAI is offering up to $445,000 (around ?3.7 crore) for a specialised research role on its Preparedness safety team. The position has drawn attention because it deals with one of the most debated questions in artificial intelligence: what happens if AI systems begin making themselves smarter without human involvement.
The role, as reported by Business Insider, centres on the idea of recursive self improvement, where an AI could design and train more advanced versions of itself with little human oversight. Researchers in this position will study speculative risks that may not exist today but could emerge quickly as AI capabilities grow.
Key responsibilities include running safety experiments on advanced models, monitoring unexpected AI behaviour, and defending systems against data poisoning attacks, where manipulated information corrupts training datasets. The job also involves developing tools to understand how AI systems reason and make decisions, and tracking how automation could start replacing technical work inside OpenAI.
The Preparedness team’s work is becoming more urgent as AI advances rapidly. Industry leaders have reportedly warned that humanity is approaching a point where AI improves itself faster than humans can keep pace. Meanwhile, it is reported that according to researchers at Model Evaluation and Threat Research or METR, the complexity of tasks AI can complete is doubling roughly every seven months, suggesting AI agents may soon handle large portions of software engineering work.
This role reflects OpenAI’s effort to anticipate and mitigate risks before they become real world problems, while offering one of the most lucrative research salaries in the industry.