A RAND–METR initiative advances safer frontier AI through rigorous risk evaluation and biosecurity research.
Photo by AscentXmedia/Getty Images
Canary is a collaboration between RAND and METR to help ensure that rapidly advancing artificial intelligence serves society safely. As AI systems grow more capable, their potential for both transformative benefit and serious harm expands. Canary brings together RAND’s deep experience in evidence-based policy research and METR’s technical expertise in evaluating autonomous AI behavior to build the science, methodology, and tools needed to understand emerging risks before new systems are released.
By developing transparent, predeployment evaluation frameworks and sharing results with the broader AI community, Canary aims to lower barriers for responsible research and contribute to a global culture of safety. The project’s work will help AI developers, policymakers, and researchers anticipate misuse and autonomy risks, guiding the evolution of frontier AI toward outcomes that protect and strengthen society.
RAND is researching the implications of AI across nearly every policy domain, but this project is exciting because it is the first to incorporate our evidence-based approach to assess the capabilities of frontier AI systems while still in development.
RAND President and CEO Jason Matheny
Canary is supported through a major philanthropic commitment from The Audacious Project, housed at TED, which has provided funding for RAND and METR to jointly pursue advances in AI safety. Additional support for related research activities is provided through RAND’s discretionary research funds and partner collaborations dedicated to the responsible development of artificial intelligence.
