Sociotechnical AI Safety Retreat at Kioloa Coastal Campus
From May 15–18 2025, researchers gathered at ANU’s Kioloa Coastal Campus for the Sociotechnical AI Safety (SAIS) Retreat – a focused, collaborative workshop on the future of AI safety research and the role of university-based institutions in shaping it.
Led by Professor Seth Lazar, the retreat brought together scholars and early-career researchers to define key terms, share new work, and lay the groundwork for future collaborations. Central to the retreat was a working definition of AI Safety as the effort to understand and mitigate severe risks posed by advanced AI systems, especially those operating with significant autonomy from human oversight.
The retreat emphasised a sociotechnical lens – understanding AI not only as software or hardware, but as part of a broader system involving people, institutions, infrastructures, and social structures. Discussions explored how this lens is essential even for “technical” AI safety, and how it can guide better research, policy, and training approaches for the age of advanced AI.
Participants included Noam Kolt, Luke Thorburn, Stephen Casper, Kevin Feng, Hannah Kirk, Max Lamparth, Alan Chan, Daniel Murfet, Liam Carroll, Tiberio Caetano, Caroline Hendy, Daniel Kilov, Seth Lazar, Jake Stone, Stevie Bergman, Lorenzo Manuali, Kevin Klyman, Taylor Sorensen, Tegan Maharaj, Kira Breithaupt, and Jennifer Munt.
The retreat created the opportunity for participants to:
Develop a conceptual and practical toolkit for sociotechnical AI safety, accessible to early-career researchers entering the field
Explore the future role of university-based institutes in AI safety
Begin work on collaborative papers and long-term projects
Build community amid the coastal bushland