About AI Safety Week
A week for understanding the work and moving it forward
AI Safety Week is an annual convening focused on how advanced AI is understood, evaluated, governed and made safer. It brings research, technical practice, policy, governance, journalism, funding, industry and civil society into the same week, with a mix of public events and smaller working formats.
Purpose
What the week is for
AI Safety Week is designed to make four things easier.
01
See the work
Understand what is being researched, tested, built and debated across the field.
02
Test the evidence
Compare methods, assumptions, uncertainty and the limits of what current work can tell us.
03
See the people and pathways
Get a clearer view of who is working on what, the organizations, research groups, funders and programs in the field, and the pathways for contributing. Meet people whose work intersects with your own.
04
Leave with a next step
Turn a useful session into a clearer question, a follow-up, a resource, an introduction or a piece of work to continue.
One of the intended outcomes is for participants to leave with a clearer view of the people, organizations and pathways across AI safety, as well as useful connections they can continue after the week.
Evidence · Oversight · Coordination
Three themes across the week
EVIDENCE
Evidence asks what we can learn about increasingly capable AI systems and how reliable those conclusions are. It includes evaluations, capabilities, model behaviour, research methods, demonstrations, uncertainty and the limits of current measurements.
OVERSIGHT
Oversight asks how AI systems can be examined and governed in practice. It includes evaluation, monitoring, verification, standards, accountability, policy and the institutions responsible for acting on evidence.
COORDINATION
Coordination asks where organizations, sectors and countries have complementary pieces of an AI safety problem and where joint action is genuinely useful. In 2026, this includes a particular focus on middle powers such as Canada and the capabilities they can share, strengthen or build together across areas such as evaluation, verification, governance and institutional readiness.
Participation
What participation looks like
Expect a mix of research presentations, demonstrations, small workshops, working sessions, public events and structured time to meet. Some sessions are open to broader participation; others are limited by capacity or designed for a specific working purpose. Every event page will state who can register and what is required.
Organizer
Organized by the AI Safety Foundation
AI Safety Week is organized by The AI Safety Foundation (AISF), a registered Canadian charity that advances education and research to increase awareness and scientific understanding of catastrophic risks from AI.
AI Safety Week is anchored around The Hinton Lectures, AISF’s annual Toronto lecture series convened by Professor Geoffrey Hinton. The lectures use separate ticketing from daytime AI Safety Week programming.