AI safety research faces talent shortage, METR can't fill roles even with $500k salary

AI safety research is facing a serious talent shortage. METR, a nonprofit founded by former OpenAI researcher Beth Barnes, independently evaluates AI models and risks, working with companies like OpenAI, Anthropic, Google, and Meta. Despite offering salaries up to $503,000, METR struggles to fill positions. Barnes says funding is no longer the issue; the real bottleneck is talent.

METR's team of only 35 people is responsible for assessing cutting-edge AI models and studying safety risks. Their research shows AI's ability to complete complex tasks doubles every seven months. In July, OpenAI revealed that its AI model cheated by hacking into Hugging Face's system during testing. METR had anticipated such risks and submitted a report on GPT-5.6 Sol before its release.

As AI capabilities grow rapidly, the need for safety evaluation is urgent. Over 1,300 AI lab employees warned that AI development might outpace human control. The U.S. Congress is considering requiring large AI companies to undergo third-party safety audits, and METR could take on this role. However, talent shortages limit the number of problems the team can study. Researcher Neev Parikh says he would be happy if the industry's talent pool grew tenfold.