Elizabeth Kelly, the inaugural director of the United States AI Safety Institute, has stepped down from her role after a year overseeing efforts to measure and counter risks from advanced AI systems. During her tenure, the institute reached agreements with OpenAI and Anthropic to test their models before release and collaborated with global AI safety organisations.

The institute, created under former President Joe Biden's administration, operates within the US Commerce Department's National Institute of Standards and Technology. Since taking office, President Donald Trump has revoked Biden's 2023 executive order on AI, raising questions about the institute's future direction under the new administration.

Kelly did not comment further on her departure but expressed optimism in a LinkedIn post, stating that the institute's mission remains crucial to the future of AI innovation. The White House has yet to clarify its plans for AI regulation and safety oversight.