Next edition in planning
AI Safety Unconference 2026
A participant-led space to make progress on AI safety.
What is AI Safety Unconference?
The AI Safety Unconference is shaped by its participants. Sessions are proposed, scheduled, and facilitated by the community to advance research on alignment, interpretability, governance, strategy, and related safety questions.
A custom scheduling tool matches multi-track discussions to attendee preferences, and shared chat and networking spaces keep collaboration going across time zones.
Last edition at a glance
In VAISU 2024, 400 participants registered, and 40 sessions were conducted. Survey respondents attended an average of 4.5 sessions and highlighted the value of new connections. Improvements for next time include stronger matchmaking and inviting more senior researchers.
A USD 2,000-equivalent prize recognized standout sessions from contributors including Steve Omohundro, Filip Sondej, Joe Rogero, Jay Luong & Nicoleta Kyosovska, Tianyi (Alex) Qiu, Remmelt Ellen, Tse Yip Fai, Sarah D'Andrea, Gergő Gáspár, and Madhusudhan Pathak.