The Defense Advanced Research Projects Agency (DARPA) is engaging industry and research organizations through a Proposers Day for its Decentralized Artificial Intelligence through Controlled Emergence (DICE) program. The event signals early movement toward a potential future solicitation focused on redefining how artificial intelligence operates in complex, contested environments.
The objective of the DICE program is to develop the underlying theory, algorithms, and architectures required to enable decentralized coordination among large numbers of AI agents. Rather than relying on centralized control or loosely connected systems, the effort seeks to create scalable, adaptive collectives of AI agents capable of executing long-duration missions while remaining aligned with human intent.
Scope focuses on advancing distributed AI systems that can dynamically organize, collaborate, and adapt in real time. These systems are designed to operate in environments where communication may be degraded, adversaries may attempt to disrupt operations, and mission conditions evolve rapidly. The program emphasizes resilience, ensuring that the loss or compromise of individual agents does not degrade overall system performance.
A central concept of the effort is “controlled emergence,” where complex system-level behavior arises from simple local rules, but remains predictable and constrained. This approach draws from principles seen in large-scale networks, enabling robust global outcomes without centralized orchestration while maintaining safeguards against unintended behaviors.
The Proposers Day is intended to provide insight into technical challenges, align industry capabilities with program goals, and facilitate teaming among potential participants.
This opportunity will likely attract AI researchers, autonomy developers, defense technology firms, and academic institutions with expertise in distributed systems, multi-agent coordination, machine learning, and resilient architectures. Organizations focused on next-generation AI systems capable of operating in contested and decentralized environments will be particularly well positioned to engage.
