An interdisciplinary research group based at Charles University in Prague studying multi-agent systems composed of humans and advanced AI, focused on understanding and mitigating systemic risks from AI integration into human institutions.
An interdisciplinary research group based at Charles University in Prague studying multi-agent systems composed of humans and advanced AI, focused on understanding and mitigating systemic risks from AI integration into human institutions.
People
Updated 05/18/26Co-founder and Principal Investigator
Researcher
Research Manager
Research Engineer
LLM Psychology and Sociology Researcher
Senior Researcher
LLM Persona Researcher
Researcher
Funding Details
Updated 05/18/26- Annual Budget
- -
- Current Runway
- -
- Funding Goal
- -
- Funding Raised to Date
- -
Org Details
Updated 05/18/26The Alignment of Complex Systems Research Group (ACS) is an interdisciplinary research group founded in January 2022 by Jan Kulveit and Tomas Gavenciak. The group is academically housed at the Center for Theoretical Study at Charles University in Prague, with administrative and fiscal support from Epistea z.s., a Czech nonprofit organization. ACS aims to develop an understanding of advanced AI systems and the complex ways in which they are embedded in and interact with humans and human institutions, contributing to making such systems safe and aligned towards shared human values including freedom, kindness, and justice. Rather than studying isolated AI agents, ACS investigates the dynamics of intelligence across multiple scales, from individual agents to complex civilization-scale systems. The group's research program spans several key areas. Their work on gradual disempowerment investigates how human influence over crucial societal systems including the economy, culture, and governance could be irreversibly eroded through incremental AI advancement, even without coordinated power-seeking by AI systems. Their AI sociology research examines interactions between diverse AI systems, human-AI teams, and existing human institutions. Their hierarchical agency work develops formalisms for understanding superagent-subagent relationships and upward and downward intentionality. They also pursue realistic models of human cognition and values, challenging classical rational agent assumptions. The team draws on expertise in physics, theoretical computer science, machine learning, cognitive science, and philosophy. Jan Kulveit, the principal investigator, holds a PhD in physics from Charles University and was previously a Research Fellow at the Future of Humanity Institute at Oxford University. Other core team members include Tomas Gavenciak (PhD in theoretical computer science), Gavin Leech (PhD candidate in machine learning, co-founder of Arb research consultancy), Simon McGregor (background in computer science, psychology, and artificial life), and Mihaly Barasz (mathematics, former Google SRE). ACS also organizes the Human-Aligned AI Summer School (HAAISS), an annual educational event held in Prague that brings together PhD students, researchers, and talented students to explore AI alignment topics. The summer school is supported by the Long-Term Future Fund and has been running since at least 2022. The group operates primarily from Prague, Czech Republic, with additional presence in London and the San Francisco Bay Area. It has received funding from the Survival and Flourishing Fund across multiple grant rounds.
Theory of Change
Updated 05/18/26ACS believes that AI safety cannot be solved by studying individual AI agents in isolation. Instead, the critical challenge is understanding how ecosystems of intelligence, composed of diverse AI systems, human-AI teams, and human institutions, evolve and interact at multiple scales. By developing foundational conceptual frameworks drawing on complex systems theory, evolutionary biology, active inference, and other disciplines, ACS aims to identify and mitigate systemic risks such as gradual disempowerment, where human influence over societal systems is incrementally eroded. Their theory of change posits that rigorous interdisciplinary research into hierarchical agency and multi-scale alignment will provide the conceptual tools needed to design governance strategies and alignment approaches that preserve human agency as AI systems become more deeply integrated into civilization.
Grants Received
Updated 05/18/26Projects
Updated 05/18/26
A grant programme in which the Alignment of Complex Systems research group in Prague accepts research proposals via acsresearch.org for projects applying complex-systems frameworks to AI alignment challenges.
A four-day summer school in Prague offering intensive talks, workshops, and discussions on AI alignment research and broader framings of AI risk for researchers and students.
Discussion
Key risk: Their complex-systems approach risks staying too abstract to yield timely, decision-relevant mechanisms for labs or policymakers, especially given a small team and limited integration with frontier developers.
Case for funding: ACS is uniquely focused on civilization-scale AI-human ecosystem dynamics—evidenced by work like Gradual Disempowerment and formal models of hierarchical agency and AI sociology—while also running HAAISS to build talent, offering high-leverage conceptual tools for governance and alignment that most agent-centric safety efforts neglect.