OpenAI is hiring an AI safety researcher for its Preparedness team with an annual salary ranging from $295,000 to $445,000. The position requires a professional who is "tasteful and strategic" to evaluate potential risks that have not yet manifested, emphasizing the importance of human judgment alongside technical coding skills. The role specifically involves defending models against data poisoning, developing tools to interpret model reasoning, and tracking internal progress regarding the automation of technical staff. These responsibilities fall under the Preparedness team's broader mandate, which includes automated red-teaming (simulated adversarial attacks), identifying biological and chemical risks, and mitigating threats posed by agentic AI (systems that can take actions to achieve goals autonomously).

The recruitment highlights the company’s focus on recursive self-improvement, where AI systems research, design, and train subsequent versions of themselves with decreasing human involvement. CEO Sam Altman has established ambitious timelines for these capabilities, aiming for an automated AI research intern running on hundreds of thousands of chips by September 2026, and a true automated AI researcher by March 2028. This shift in capability is supported by industry observations; for instance, research from METR indicates that the task length frontier models can handle doubles every seven months. Additionally, Anthropic policy head Jack Clark estimates the probability of AI conducting autonomous R&D without human intervention by the end of 2028 at approximately 60 percent. Meanwhile, Demis Hassabis of Google DeepMind has described the current state of AI progress as being at the foothills of the singularity, a theoretical point where AI growth becomes uncontrollable and irreversible.

OpenAI is hiring an AI safety researcher for its Preparedness team with an annual salary ranging from $295,000 to $445,000. The position requires a professional who is "tasteful and strategic" to evaluate potential risks that have not yet manifested, emphasizing the importance of human judgment alongside technical coding skills. The role specifically involves defending models against data poisoning, developing tools to interpret model reasoning, and tracking internal progress regarding the automation of technical staff. These responsibilities fall under the Preparedness team's broader mandate, which includes automated red-teaming (simulated adversarial attacks), identifying biological and chemical risks, and mitigating threats posed by agentic AI (systems that can take actions to achieve goals autonomously).

The recruitment highlights the company’s focus on recursive self-improvement, where AI systems research, design, and train subsequent versions of themselves with decreasing human involvement. CEO Sam Altman has established ambitious timelines for these capabilities, aiming for an automated AI research intern running on hundreds of thousands of chips by September 2026, and a true automated AI researcher by March 2028. This shift in capability is supported by industry observations; for instance, research from METR indicates that the task length frontier models can handle doubles every seven months. Additionally, Anthropic policy head Jack Clark estimates the probability of AI conducting autonomous R&D without human intervention by the end of 2028 at approximately 60 percent. Meanwhile, Demis Hassabis of Google DeepMind has described the current state of AI progress as being at the foothills of the singularity, a theoretical point where AI growth becomes uncontrollable and irreversible.