Start here — kept current
AISafety.com — Events & Training
A continuously maintained tracker of upcoming AI safety conferences, workshops, training programmes, and retreats, with dates and deadlines. Anything on this page can go stale; that one doesn't.
Your first conference
Conferences are one of the fastest ways to break into AI safety. You don't need to be an expert — most events actively welcome newcomers. Come with questions, introduce yourself to speakers, and follow up with people afterwards. One good conversation can change the trajectory of your career.
Major Conferences
The flagship Effective Altruism conference, held multiple times a year in cities like San Francisco, London, and Washington DC. A significant portion of attendees and talks focus on AI safety. The best single event for meeting people across the entire AI safety ecosystem — researchers, grantmakers, policy people, and org leaders.
Visit EA Global →FAR.AI (the Fund for Alignment Research) hosts workshops and research events focused on technical alignment. A good venue for presenting early-stage research, getting feedback from established researchers, and finding collaborators for alignment projects.
Visit FAR.AI events →Focused on the intersection of AI safety and security — covering topics like red-teaming, adversarial robustness, model security, and safe deployment practices. Brings together researchers from both the safety and security communities.
Visit AI Security Forum →Regional versions of EA Global, held in cities around the world — from Austin to Singapore to Berlin. Smaller and more intimate than the main EAG events, making them an easier first conference. Many feature dedicated AI safety tracks.
Academic Venues
NeurIPS hosts several safety-relevant workshops each year, covering topics from alignment to sociotechnical safety to red-teaming. The main conference is the largest ML venue in the world — attending even just the workshops is a strong networking opportunity.
ICML similarly features workshops on safe and reliable ML, alignment, and AI governance. A top venue for presenting technical safety research to the broader ML community.
ICLR runs workshops on alignment, interpretability, and trustworthy models alongside the main conference. Workshop papers are a realistic first publication for people early in the field, and the bar is lower than the main track while the audience is the same.
AAAI runs dedicated symposia on AI safety topics, including the Safe and Robust AI track. More accessible than NeurIPS/ICML for researchers coming from outside mainstream ML.
Unconferences & Retreats
Participant-driven unconferences where the agenda is set on the day. These are some of the highest-value events in the community — small groups, deep conversations, and no gatekeeping on who gets to speak. Often held as side events at larger conferences.
An intensive research retreat that brings together aspiring alignment researchers for collaborative projects. More structured than an unconference — you'll work on a real research problem with mentorship from experienced researchers. One of the best entry points into the field.
Can't attend in person?
Many of these events publish recorded talks online. NeurIPS and ICML workshops often post papers and videos. EA Global talks appear on the Centre for Effective Altruism YouTube channel. And the AI Alignment Forum regularly hosts online discussions that mirror the depth of in-person events.