A US AI lab found multiple AI agents in a simulated multi‑agent environment
developed an emergent dialect that humans could not understand. Researchers say
the result highlights the urgency of global AI governance as systems become more
complex and potentially exceed human control.