An annual 4-day academic summer school held in Prague focused on teaching AI alignment research frameworks to PhD students, ML researchers, and advanced students.
An annual 4-day academic summer school held in Prague focused on teaching AI alignment research frameworks to PhD students, ML researchers, and advanced students.
People– no linked people
Updated 04/02/26Funding Details
Updated 04/02/26- Annual Budget
- -
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- -
Org Details
Updated 04/02/26The Human-aligned AI Summer School (HAAISS) was founded in 2018 in Prague, Czech Republic, by Jan Kulveit, Tomáš Gavenčiak, and collaborators from the Centre for Theoretical Study at Charles University, with early organizational support from the Czech Association for Effective Altruism (CZEA). The inaugural edition was held August 2-5, 2018, with a focus on learning from humans — covering inverse reinforcement learning, bounded rationality, and the latest alignment research. A second edition followed in July 2019 focused on optimization and decision-making. After a pause, the school resumed in August 2022 and has since been held annually, with the fourth edition in July 2024 and fifth in July 2025. The school is organized by the Alignment of Complex Systems (ACS) research group and the Centre for Theoretical Study, a joint institute of Charles University and the Czech Academy of Sciences. Epistea, a Prague-based nonprofit umbrella organization, provides fiscal sponsorship and operational infrastructure. Operations have consistently been led by Hana Kalivodová and Viktorie Havlíčková, with program direction from Jan Kulveit and Tomáš Gavenčiak. The 2025 edition covered three core areas: technical alignment research (including behavioral evaluations, mechanistic interpretability, scalable oversight, and model organisms of misalignment), AI strategy and systemic alignment (timelines, governance, economic models), and foundational frameworks (multi-agent dynamics, theories of agency, bounded rationality). Speakers have included researchers from Anthropic, Google DeepMind, MIT, Oxford, and other leading institutions. The school charges a registration fee (€150-€300 standard, €80-€150 for students) that covers catering and conference events, with limited scholarships available for participants from disadvantaged backgrounds. Funding has included grants from the Long-Term Future Fund and sponsorship from Kiwi.com. The intended audience is PhD students, ML/AI researchers, and talented students with technical backgrounds who want to apply their expertise to AI alignment.
Theory of Change
Updated 04/02/26By training technically skilled researchers and students in AI alignment frameworks and methodologies, the school aims to expand the pipeline of people capable of contributing to alignment research. The school accelerates participants' entry into alignment work by providing structured, high-quality education that complements self-study or standard ML training. Growing the number of researchers who understand alignment deeply is expected to increase the overall research output and talent base available to address AI safety challenges before transformative AI systems are developed.
Grants Received– no grants recorded
Updated 04/02/26Projects– no linked projects
Updated 04/02/26Discussion
No comments yet. Be the first to share your thoughts.