OpenAI Offers Up to $445K for New AI Safety Job Amid Push to Tackle Self-Improving AI

OpenAI is offering as much as $445,000 for a newly listed artificial intelligence (AI) role focused on one of the sector’s most debated risks, the possibility that AI systems could eventually improve themselves.

The position, part of the company’s Preparedness safety team, is aimed at helping OpenAI prepare for what researchers call “recursive self-improvement,” where AI tools could begin researching and designing more advanced versions of themselves.

According to a report by Business Insider, the role comes with a salary range of $295,000 to $445,000 and seeks researchers to tackle future-facing technical risks linked to increasingly powerful AI systems.

Inside OpenAI’s High-Paying New AI Safety Role

The listing seeks “strong technical executors to support preparations for recursive self-improvement,” indicating that OpenAI is prioritising safety research alongside rapid advances in model development.

“This work relies on reasoning about problems that might exist in the future, but might not exist now,” the listing says, adding that “so it’s especially important that people in this role are tasteful and strategic”.

The researcher would work within OpenAI’s Preparedness team, which focuses on anticipating severe AI-related risks.

According to the listing, responsibilities may include defending models from data-poisoning attacks, building tools to interpret how AI systems reason, and testing safeguards for increasingly autonomous systems.

The role may also involve tracking “progress toward automation of technical staff,” reflecting growing industry interest in how AI could eventually handle software engineering and research work currently done by humans.

AI Labs Race to Prepare for Self-Improving Systems

The hiring effort comes as frontier AI capabilities continue advancing at an unexpected pace. Researchers at the METR lab wrote in March that the length of tasks cutting-edge AI models can complete appears to double roughly every seven months, a trend that could rapidly expand what such systems can accomplish.

OpenAI has openly discussed ambitions to automate parts of AI research itself. Chief Executive Sam Altman said last year that the company hopes to develop an automated AI research intern eventually and later a more advanced “true automated AI researcher”.

“We may totally fail at this goal,” Altman wrote on X, “but given the extraordinary potential impacts, we think it is in the public interest to be transparent about this.”

The unusually high pay package also underscores an escalating talent battle across the AI sector, where companies are competing aggressively for researchers with expertise in alignment, model safety and long-term AI risks.