$6,000 a Month AI Alignment Fellowship (February 2027)
- Organization: Iliad
- Type: Fellowship
- Location: Global
- Deadline: December 16, 2026
- Age: 18 and over
- Funding: Fully Funded
Iliad, a non-profit AI safety research organisation based in London and the San Francisco Bay Area, runs a fully funded, in-person research fellowship that helps researchers from mathematics, theoretical physics and computer science move into technical AI alignment. Fellows spend three months doing full-time, mentored research, after first completing Iliad's month-long Intensive course. The February 2027 cohort runs from 1 February to 23 April 2027 in London or the Bay Area. **What you get:** - A $6,000 USD travel-and-housing allowance for each month of the Fellowship - A research mentor and a place in a community of alignment researchers - No tuition for the Intensive, which also pays a $5,000 USD travel-and-housing allowance, office space, and lunch and dinner five days a week - A chance at a six-month funded extension at the same $6,000 a month, offered to about half of Fellows **What you'll do:** - Complete the Iliad Intensive, a four-week, full-time taught course in foundational AI alignment, before the research months (the January 2027 Intensive in the Bay Area is the one that precedes this cohort) - Spend three months on full-time, individual mentored research in AI alignment, in person in London or the Bay Area - Explore a research direction in areas such as learning theory, interpretability, agency or safety guarantees **Key dates:** - Applications close 16 December 2026 - February 2027 Fellowship runs 1 February to 23 April 2027, London or Bay Area - The only Intensive on Iliad's calendar before then is the January 2027 Intensive, 4 to 29 January 2027 in the Bay Area, with applications closing 16 November 2026 **More information:** - One Iliad application form covers the upcoming Intensive and Fellowship cohorts; tick the cohorts you want - Questions go to contact@iliad.ac
Apply for $6,000 a Month AI Alignment Fellowship (February 2027)

