Virtual Agent Economies
Nenad Tomasev, Matija Franklin, Joel Z. Leibo, Julian Jacobs, William A. Cunningham, Iason Gabriel, Simon Osindero
2025-09-15
Summary
This paper explores the idea that as AI becomes more advanced, we're starting to see a new kind of economy emerge – one where AI agents are buying, selling, and working with each other, often without direct human control.
What's the problem?
The core issue is that this AI-driven economy is developing quickly and somewhat unpredictably. It could lead to really efficient systems, but also carries risks like economic instability and widening the gap between the rich and poor. We don't fully understand how this 'sandbox economy' will interact with the existing human economy, or how to control it to prevent negative consequences.
What's the solution?
The researchers propose thinking about this AI economy based on how it started (naturally or by design) and how separate it is from our current economy. They then suggest ways to build this AI economy more intentionally, focusing on things like fair auction systems for resources, designing AI systems that work towards shared goals, and creating infrastructure to build trust and accountability within the system.
Why it matters?
This research is important because AI is rapidly changing the world, and this new AI economy has the potential to be incredibly powerful. By proactively designing how these AI agents interact, we can try to steer this change in a way that benefits everyone, rather than letting it develop chaotically and potentially cause harm.
Abstract
The rapid adoption of autonomous AI agents is giving rise to a new economic layer where agents transact and coordinate at scales and speeds beyond direct human oversight. We propose the "sandbox economy" as a framework for analyzing this emergent system, characterizing it along two key dimensions: its origins (emergent vs. intentional) and its degree of separateness from the established human economy (permeable vs. impermeable). Our current trajectory points toward a spontaneous emergence of a vast and highly permeable AI agent economy, presenting us with opportunities for an unprecedented degree of coordination as well as significant challenges, including systemic economic risk and exacerbated inequality. Here we discuss a number of possible design choices that may lead to safely steerable AI agent markets. In particular, we consider auction mechanisms for fair resource allocation and preference resolution, the design of AI "mission economies" to coordinate around achieving collective goals, and socio-technical infrastructure needed to ensure trust, safety, and accountability. By doing this, we argue for the proactive design of steerable agent markets to ensure the coming technological shift aligns with humanity's long-term collective flourishing.