Research paper identifies open problems in frontier AI risk management across planning, identification, analysis, evaluation, and mitigation stages

A systematic review by Ziosi, Plueckebaum, Casper, and 15 other authors examines unresolved challenges in frontier AI risk management, surfacing open problems at each stage of the risk management process.

The work classifies open problems into three categories: lack of scientific or technical consensus, misalignment with established risk management frameworks, and implementation shortcomings despite apparent consensus.

The paper identifies developers, deployers, regulators, standards bodies, researchers, and third-party evaluators as actors positioned to address frontier AI risk challenges, and is accompanied by a living online repository intended to support coordination and guide future research and governance efforts.