AI Alignment Forum

The AI Alignment Forum is the central online venue for technical AI alignment research discussion. It was officially launched on October 29, 2018, following a two-month open beta announced in July 2018. The forum was developed by the LessWrong 2.0 team — led by Oliver Habryka, Ben Pace, Raymond Arnold, and Jim Babcock — and superseded MIRI's Intelligent Agent Foundations Forum, whose content was migrated to the new platform. The forum is operated by Lightcone Infrastructure, a Berkeley-based nonprofit that also runs LessWrong.com and Lighthaven (a community event venue). Researchers from major AI safety organizations including MIRI, OpenAI, DeepMind, CHAI, FHI, and Open Philanthropy participate in forum discussions. Content posted on the Alignment Forum automatically cross-posts to LessWrong, enabling a broader audience while maintaining a researcher-focused space on the forum itself. The forum launched with three foundational sequences on Embedded Agency, Iterated Amplification, and Value Learning to help onboard new researchers. The platform prioritizes high epistemic standards and is widely regarded as one of the most important coordination and knowledge-sharing venues in the AI safety field. Lightcone Infrastructure, which operates the forum, transitioned from approximately $8M/year in total operations to roughly $3M/year in 2023 following changes in the EA funding landscape. The LessWrong and Alignment Forum software team accounts for approximately $1.7M/year of that budget. Original funders included BERI, EA Grants, Nick Beckstead, and Matt Wage; Open Philanthropy later became the major funder with grants totaling approximately $4.5M in 2023. The organization is actively fundraising in 2026, with a goal of $3M to sustain operations.