Agentic and Generative AI as Artificial Life

Agentic and Generative AI as Artificial Life
Complexity, Safety, and Security in Ecologies of Autonomous Systems

at ALIFE 2026

Conference Dates: August 17-21, 2026 - Waterloo, Ontario, Canada

Aim of the workshop

AI systems are rapidly evolving from passive, prompt-driven interfaces into persistent, tool-using, self-directing agents that plan, act, coordinate, and modify their own capabilities over time. Emerging platforms inspired by systems such as OpenClaw and MoltBook exemplify this transition: they enable agents to operate across applications, maintain memory, acquire and reuse skills, and participate in growing interconnected "skill ecosystems".

In parallel, generative models such as large language models (LLMs) are introducing unprecedented levels of semantic and cognitive richness, enabling agents not only to act autonomously, but to serve as sources of complexity in their own right.

Together, these developments mark a shift toward artificial systems that are increasingly ecological, developmental, and open-ended -and raise a foundational question for Artificial Life: when does engineered autonomy begin to resemble a living system?

This workshop positions agentic AI infrastructures and generative AI capabilities as a new experimental substrate for ALife research. AI agents are not merely intended as software components, but as organisms embedded in digital environments -competing for resources (attention, compute, permissions), forming symbiotic relationships with tools and humans, adapting through memory and feedback, and generating emergent collective dynamics.

Crucially, today's agent ecosystems are not hypothetical. Skill marketplaces, tool-use frameworks, and cross-platform orchestration systems already define ecological niches and interaction "physics." Within these environments:

tool-use and API access define metabolic capabilities

memory enables developmental trajectories and identity formation

skill sharing and reuse create pathways for cultural transmission

prompt injection and tool misuse resemble parasitism and predation

orchestration frameworks structure interaction, coordination, and constraint

At the same time, generative AI fundamentally reshapes the ALife paradigm. Classical ALife has focused on how complexity emerges from simple rules; LLM enable investigating how non-trivial structures emerge from agents that already possess rich internal representations and linguistic capabilities. Examples include: collective innovation and cultural evolution in LLM agent groups, emergence of social norms through language-mediated interaction, individuality and norm differentiation from homogeneous populations, evolution of linguistic traits as ecological variables, collective biases arising from individually unbiased agents, reinterpretations of multi-agent systems as unified cognitive entities.

The same properties that make these ecosystems adaptive and powerful also introduce critical safety, security, and governance challenges. Agentic systems increasingly delegate tasks across chains of agents and tools, operate across organizational and technical boundaries, modify their own behavior through memory and skill acquisition, interact in decentralized networks.

This leads to accountability diffusion, where responsibility for outcomes (e.g., harmful outputs, resource misuse, cascading failures) becomes difficult to attribute. Unlike natural ecosystems, however, these engineered systems are expected to be governable and auditable. As a result, safety and alignment must be reframed as emergent regulatory processes within complex adaptive systems.

Mechanisms such as reputation, trust, and norm formation may play roles analogous to ecological regulation -but their effectiveness depends on maintaining sufficient legibility of agency and causation within increasingly complex interaction networks. This tension between open-ended autonomy and enforceable governance is a central design challenge for next-generation AI systems.

This workshop aims to bridge ALife, agentic AI, and generative AI communities by welcoming contributions spanning theoretical, empirical, and experimental work on:

Agentic Ecosystems as Digital Ecologies
Generative AI as a Source of Complexity
Safety, Security, and Emergent Governance

Rather than asking whether these systems are "alive", this workshop explores how Artificial Life can contribute to the design, analysis, and governance of increasingly autonomous generative AI ecosystems -and how these systems, in turn, expand the empirical and theoretical scope of ALife.