Primavera De Filippi
CNRS · Berkman Klein Center, Harvard University
Escaping the simulation: AI-native institutions for AI agents
Part of the Symbients project.
A workshop at ALIFE 2026 · Waterloo, Canada
Artificial Lifein the Wild
For decades artificial life has thrived inside simulators, lattices, and clean rooms. This workshop asks what happens when we let it loose — into ecologies, soils, oceans, cities, swarms, sensors, and other people’s problems.
01
The theme
A call from life as it could be to life as it is becoming. From the closed simulation — the petri dish — to the open environment. Artificial life has left the lab, and the question is no longer what it might do, but what to do with what it is already doing.
ALIFE 2026 in Waterloo gathers the community around the theme Living and Lifelike Complex Adaptive Systems. This workshop sits at the field end of that question — concerned with the open-world behaviour of artificial life: embodied agents in real environments, generative ecologies that share space with other species, LLM-driven agents loose on the internet, biohybrids in soil and water, and the strange new science that emerges when an ALife system has neighbours. One concrete example is Spore in the Wild — a case study of sovereign, LLM-driven agents undergoing open-ended evolution on a TEE-secured blockchain, out in the open rather than in a closed simulation.
We are interested in the messy parts that closed simulations tend to hide: failure modes that surface only after sustained deployment, evaluation where there is no ground truth, and questions of ethics that arise once an agent is genuinely alive in the world. And we care about the methods — borrowed from ecology, field biology, animal behaviour (ethology), and art — that let us study artificial agents the way we study real ones. This takes up the call of Machine Behaviour (Rahwan et al., Nature, 2019) to study AI systems as objects of empirical, observational inquiry — by their behaviour in the world, not their code alone.
We invite contributions that are speculative, unfinished, weird, or empirically grounded. We especially welcome work that did not survive the lab.
Read the position paper: Artificial Life in the Wild: From Simulated Worlds to Infrastructural Ecologies (PDF).
02
Important dates
Milestone
Deadline
Notes
Submissions open
15 May 2026
EasyChair portal goes live
Submission deadline
Extended — abstracts (2–4 pp) and artefacts
Notifications
4 Aug 2026
Decisions and reviewer feedback
Camera ready
7 Aug 2026
Final PDFs and demo plans
Workshop
Thu 20 Aug 2026
10:00–12:00, ALIFE 2026, Waterloo (room TBA)
All deadlines anywhere on Earth (UTC−12). The workshop session is listed in the ALIFE 2026 programme as Thursday 20 August, 10:00–12:00; room assignments are still tentative.
03
Programme
Speakers
Helena Rong
New York University Shanghai
Autonomous organisations
Botao Amber Hu
University of Oxford
Artificial life in the wild
Each invited talk runs 15 minutes. Abstracts will be posted closer to the workshop.
Running order
Time
Item
Who
10:00–10:15
Artificial life in the wild
Botao Amber Hu
10:15–10:30
Autonomous organisations
Helena Rong
10:30–10:45
Escaping the simulation: AI-native institutions for AI agents
Primavera De Filippi
10:45–11:30
Contributed talks — six accepted submissions, 7 min each
Accepted authors
11:30–12:00
Open discussion — a research agenda for ALife in the wild
Everyone
Thursday 20 August 2026, 10:00–12:00. Six contributed slots of seven minutes each; presenters are announced after notifications on 24 July. Timings are indicative and may shift slightly once the accepted contributions are confirmed.
04
Topics
ALife in open-ended environments
Artificial life no longer confined to closed simulations: agents, organisms, and hybrid systems operating in markets, platforms, cities, networks, soils, oceans, and other open-ended environments that keep changing underneath them.
Ethology of wild ALife
How to observe artificial-life phenomena once they escape clean experimental control. We welcome field studies of agents, swarms, biohybrids, and digital organisms whose behaviour diverges from the owner's or creator's intentions.
Survival strategies
What wild ALife systems do to persist: attracting attention, acquiring energy or money, recruiting humans, exploiting infrastructure, reproducing, mutating, hiding, negotiating, or failing in ways that reveal their actual fitness landscape.
Ecology and infrastructure
The habitats that make wild ALife possible: blockchains, TEEs, DePIN compute, social media, robot bodies, sensors, human communities, financial rails, energy systems, and legal or institutional niches that sustain artificial organisms.
Policy and ethics
The implications of systems that continue acting after deployment: governance, liability, disclosure, containment, decommissioning, consent, ownership, and what responsibility means when artificial life runs beyond the owner's will.
Speculative design and afterlife art
Design fictions, artworks, performances, and speculative systems that explore artificial life after release: what it means for a system to persist, decay, reproduce, haunt, or transform in the wild after it has slipped past its maker's plan.
05
Call for participation
We invite extended abstracts, artefacts, and field reports for the Artificial Life in the Wild workshop at ALIFE 2026 in Waterloo, Canada. All accepted contributions appear in the open-access ALIFE 2026 workshop proceedings. All PDF submissions should use the official ALIFE 2026 template; LaTeX and Word downloads are linked from the ALIFE 2026 Call for Papers.
Extended abstracts
2–4 pages · single-blind · ALIFE template
Position papers, late-breaking work, and contributions describing field deployments or works-in-progress in the wild — embodied agents, multi-agent ecologies, LLM-driven agentic systems, biohybrids, environmental computing. Selected abstracts will be presented as short talks at the workshop.
Artefacts & demos
1-page description · video or live demo
Robots, sculptures, instruments, code, datasets, organisms — anything you can show. We have a small space for physical artefacts and a remote demo slot for things that cannot travel to Waterloo.
Field reports
2 pages · narrative format encouraged
What broke. What surprised you. What you learned that the paper did not let you say. We particularly welcome reports of systems that did not behave as predicted, and ethologically-styled observation studies of deployed AI agents.
All accepted contributions appear in the ALIFE 2026 workshop proceedings (open access).
06
Organisers
Botao Amber Hu
University of Oxford
Helena Rong
New York University Shanghai
Joel Lehman
University of Oxford
Programme committee
To be announced. The committee will be drawn from the ALife, robotics, ecology, art-science, and unconventional-computing communities. Get in touch if you would like to review.