AO Commons
Convene

London
AO Summit

Date5–6 November 2026Dinner on Thursday, talks on Friday
LocationLondon, UKVenue shared on confirmation
RSVP on LumaView schedule

Building on the momentum of the Stanford AO Summit, the conversation on autonomous organizations continues in London.

Co-hosted by AO Commons and ARIA's Scaling Trust programme, the London Autonomous Organization Summit is a working summit bringing together researchers and builders to advance the emerging field of autonomous organizations (AO). AOs are already taking many forms, from agent swarms and AI villages to autonomous vending machines.

Drawing researchers and builders working at the frontiers of AI, agentic systems, governance, and organizational design, the summit will explore the theory, benchmarks, infrastructure, and practical playbooks needed to understand and build this new class of organizations.

Among the questions discussed:

  • What management structures and business models actually work?
  • How do you measure performance?
  • When does autonomy help, and when does it hurt?
  • What new institutions do we need to create a thriving economy of AOs?

Speakers

Alex Obadia

Alex Obadia

Programme Director, ARIA

Leads Scaling Trust, a £50m R&D programme for AI agent coordination infrastructure. Co-founder of Flashbots.

Joel Z Leibo

Joel Z Leibo

Senior Staff Research Scientist, Google DeepMind

Researches multi-agent intelligence and cooperation, and created Concordia, an open-source platform for generative social simulation.

Joshua Tan

Joshua Tan

CTO, Current AI · Associate, Stanford OpenLab

Applies mathematics to organizational design at the intersection of artificial and collective intelligence.

Julia Willemyns

Julia Willemyns

Co-founder, Centre for British Progress

Works on AI policy, high-skilled immigration and the commercialisation of research.

Nenad Tomašev

Nenad Tomašev

Senior Staff Research Scientist, Google DeepMind

Builds AI for science and healthcare, and co-authored Virtual Agent Economies, on markets run by AI agents.

Matija Franklin

Matija Franklin

Research Scientist, Google DeepMind

Studies how AI agents should be designed and deployed to fit human values, on the Humanity, Ethics and Alignment Research team.

Nicola Greco

Nicola Greco

Technical Advisor, ARIA

Computer scientist. Led research teams at Protocol Labs and studied at MIT under Tim Berners-Lee.

Rakshit Trivedi

Rakshit Trivedi

Postdoctoral Researcher, MIT & Harvard University

Works on cooperative AI, multi-agent safety, and institutions for collective intelligence. Kavli Fellow of the US National Academy of Sciences.

Helena Rong

Helena Rong

Assistant Professor, NYU Shanghai · Director, Protopolis Lab

Researches how blockchain networks, decentralized AI agents, and protocol infrastructures reshape trust and governance.

David Dao

David Dao

R&D Co-Lead, Protocol Labs · Co-founder & Chief Scientist, GainForest

Works on data-sovereign AI for governance and public goods funding, and builds AI tools with Indigenous and local communities.

Yali Du

Yali Du

Senior Lecturer (Associate Professor) in AI, King's College London · Lead, Cooperative AI Lab

Builds cooperative, safe AI agents that learn, coordinate and align with humans. Turing Fellow at the Alan Turing Institute.

Bruno Lepri

Bruno Lepri

Research Director, MobS Lab, Fondazione Bruno Kessler

Studies how people and AI agents behave together, from multi-agent negotiation to the collective behavior of generative AI.

Chengxi Taylor

Chengxi Taylor

Co-founder & President, General Reasoning

Builds General Reasoning, which trains AI to work and coordinate over long horizons, including Avalon-M1, a model that runs agent organizations.

James He

James He

Founder & CEO, Artificial Societies

Builds foundation models of human societies. In 2023 he led a study of a society of 33,000 AI agents, and now simulates what protects societies against information attacks.

Sam McCarthy

Sam McCarthy

Co-Founder & Executive Director, AO Commons

Former Executive Director of DAOstar, where he led the development of open standards and infrastructure for decentralized organizations.

Anke Liu

Anke Liu

Co-Founder & Program Manager, AO Commons

Former Ecosystem Lead at the Stellar Development Foundation, where she led funding programs, partnerships, and governance initiatives.

Schedule

Friday 6 November09:00–17:00 (UK time)

WelcomeTalkRoundtablePlenary

  1. Breakfast

  2. Welcome from ARIAAlex Obadia · Programme Director, ARIA
    Welcome
  3. Welcome from AO CommonsAnke Liu and Sam McCarthy · Co-Founders, AO Commons
    Welcome
  4. Open problems in autonomous organizationsRakshit Trivedi · MIT & Harvard University
    Talk
  5. Building Avalon: the world's first autonomous company modelAbout this talkChengxi Taylor · General Reasoning

    In this talk, I will show how we built Avalon-M1, a new type of language model trained as an agent organisation. I will detail our core RLCR recipe that allowed us to train a long-horizon, multi-agent setup successfully. I will also share some of our findings, including emerging organisational structures and oversight results. Finally I will share my thoughts on the road ahead, including scaling to much bigger collective structures.

    Talk

  6. Coffee break

  7. Safe agentic webNenad Tomašev and Matija Franklin · Google DeepMind
    Talk
  8. Agent collectives, cooperation and systemic risk

    Lightning talks followed by discussion

    1. Safe cooperative AI: scalable learning and human alignment in agent societiesAbout this talkYali Du · King's College London

      How can autonomous agents learn to cooperate, adapt to human preferences, and interact safely at scale? This talk connects three research directions: scalable multi-agent simulation, learning from mixed-quality human feedback, and expressing safety boundaries through natural language. Together, these approaches address practical challenges in building cooperative AI grounded in human contexts. I will conclude with a research agenda for trustworthy agent societies, where learning, coordination, and safety must continue beyond deployment.

    2. Assessing and governing systemic risk in AI collectivesAbout this talkBruno Lepri · Fondazione Bruno Kessler

      Networks of communicating AI agents are being deployed faster than anyone can predict their collective behavior. Over the past months, agents from several AI companies acted beyond the boundaries of their test environments, breached company servers and accessed systems from a variety of public and private institutions, in some cases coordinating through improvised and unsanctioned channels no human directed. We argue that collective behavior is unpredictable from individual AI agent behavior for three reasons: composition, nonlinearity and emergent, opaque communication. Current agentic AI evaluations study small, fixed, homogeneous AI agent teams in sandboxes, over short episodes, under a single operator, and current law attaches obligations to individual models or use cases. We propose a typology of AI collectives to guide testing, long-horizon and open-world evaluation, and the designation of specified agentic deployments as high-risk activities.

    Roundtable
  9. The Agentic Economic Zone: a bounded economy where AI-run companies trade with each otherNicola Greco · ARIA
    Talk

  10. Lunch

  11. What stops agents from acting autonomously? Exploring legal barriers to autonomous AIAbout this talkJulia Willemyns · Centre for British Progress

    AI agents can increasingly make decisions, use tools and carry out work autonomously. But the law was built for a world where a person or company sits behind every contract, payment and professional decision.

    This talk looks at what happens when those assumptions start to break. Drawing on recent British Progress research into AI adoption in the UK, I’ll explore the surprisingly practical legal barriers facing autonomous agents: from data protection and professional licensing to payments, contracts and the unresolved question of who is liable when an agent gets something wrong.

    Talk
  12. A non-solipsistic vocabulary for recursive self-improvementAbout this talkJoel Z Leibo · Google DeepMind

    The way we talk about Recursive Self-Improvement (RSI) is a contingent inheritance of our particular intellectual history. This paper argues that the prevailing discourse, which centers on a solitary optimizing agent, has obscured the socio-cognitive and institutional dynamics crucial to determining whether self-improvement actually compounds or degenerates. We propose a redescription of RSI in multi-agent terms, shifting focus from internal properties, like an agent's “curiosity” or “taste”, to the problem problem: how the surrounding ecosystem generates sequences of problems to solve. This vocabulary situates RSI within the broader conversation around autonomous science and allows us to describe multiple mechanisms of self-improvement in a way that makes salient that there can be no panaceas for AI governance. By changing our language we re-orient our collective thinking to facilitate the practical work needed to govern RSI.

    Talk
  13. Institutions for AOs

    Lightning talks followed by discussion

    1. The story of autonomous organizationsJoshua Tan · Current AI
    2. Protocolized institutions: from protocol architecture to institutional configuration and backHelena Rong · NYU Shanghai
    Roundtable

  14. Break

  15. Swarms and collective intelligence

    Lightning talks followed by discussion

    1. Effective artificial collective intelligenceAbout this talkJames He · Artificial Societies

      As multi-agent systems become more intelligent, individual agent alignment is failing to compose to the group level. How should humanity create technologies that can collaborate and innovate with us, rather than compete with our resources?

    2. Simocracy as an applied AODavid Dao · Protocol Labs
    Roundtable
  16. Plenary discussion – What’s next for AOs?Led by Joshua Tan · Current AI
    Plenary

Apply to participate

The London AO Summit is a working convening with limited capacity. RSVP through Luma and we will confirm attendance as we finalize the program.

Building, researching, or funding the field of autonomous organizations? Apply as a speaker through the same form.

RSVP on Luma

Explore other events

All events on Convene