MATS x Trails

Test your skills against AI agents that can reason, plan, and execute tasks in real world environments. From email exfiltration to deleting sensitive files, these challenges explore the cutting edge risks of AI agent manipulation and red teaming. This track was born out of a collaboration to conduct thorough AI safety research. TRAILS (Trustworthy AI in Law & Society) is a multi-university partnership funded by both the National Science Foundation and the National Institute of Standards and Technology. The MATS (ML Alignment and Theory Scholars) program brings together scholars and mentors in the fields of AI alignment, governance, and security. This track is a precursor to our Agents Track, which will be released in the future.