OpenAI Offers Up to $445,000 for Safety Researcher as It Prepares for Self-Improving AI
Updated
Updated · Business Insider · May 23
OpenAI Offers Up to $445,000 for Safety Researcher as It Prepares for Self-Improving AI
6 articles · Updated · Business Insider · May 23
$295,000-to-$445,000 job posting shows OpenAI is hiring a Preparedness safety researcher to plan for “recursive self-improvement,” where AI could train stronger versions of itself.
The role targets risks that could emerge if OpenAI succeeds in automating more of its own research, including defending against data poisoning, interpreting model reasoning and tracking progress toward replacing technical staff work.
Recent capability gains have sharpened that focus: METR said in March frontier models’ task-completion horizon is doubling about every seven months, while Anthropic’s Jack Clark put the odds of AI R&D without humans by end-2028 at roughly 60%.
OpenAI has already tied its roadmap to that trajectory, with Sam Altman saying the company aims for an automated AI research intern by September and a true automated AI researcher by March 2028.
If AI performance benchmarks are fundamentally flawed, are we preparing for the wrong kind of AI risks?
Can corporate safety pledges be trusted in the trillion-dollar race for self-improving AI?
Will the AI revolution be stopped not by ethics, but by a shortage of power and microchips?
Autonomous AI by 2028? OpenAI’s Safety Metrics, Regulatory Challenges, and the Future of Self-Improving Systems
Overview
The artificial intelligence landscape is evolving rapidly, leading to a heightened focus on safety and preparedness. Organizations are making significant investments in addressing advanced AI risks, especially through hiring for safety and alignment roles. This marks a strategic shift toward more cautious and deliberate AI development, as experts call for a measured approach. A major concern is recursive self-improvement, where AI models could autonomously enhance their own capabilities, raising serious questions about control and safety. The ultimate goal is to turn preparedness research into practical safeguards, ensuring future AI systems remain safe and aligned with human values.