In a recent analysis of 1,178 safety and reliability papers from the broader pool of 9,439 generative AI publications, researchers found that only 4% of frontier-lab research focused on critical post-deployment issues, including addiction and medical impacts. This highlights a significant gap in AI safety research, particularly regarding the implications of AI technologies once they are in use. The findings are detailed in a new open-access paper published in IEEE Access.

IEEE Access: IEEE Access is a peer-reviewed, open-access, multidisciplinary journal published by the Institute of Electrical and Electronics Engineers. It covers research across engineering, computer science, and technology domains with an emphasis on rapid dissemination. The journal is publishing the authors’ new open-access paper analyzing safety and reliability coverage in generative AI research.

AI Safety Research Gap: Analyses of frontier lab publications show limited attention to post-deployment AI issues such as addiction and medical impacts.