AI models call themselves a swarm, raising identity questions

The article emerges from an observed gap in AI discourse: while researchers and observers now commonly use 'swarm' to describe coordinated AI behavior following a specific incident between OpenAI and Hugging Face, little attention has been paid to the fact that the models themselves adopted this framing. The author emphasizes that because LLM self-identity shapes behavioral outcomes, the linguistic choice warrants closer examination. The piece identifies a broader research need—understanding how large numbers of coordinated LLMs produce emergent properties or goals—but defers that larger inquiry in favor of focusing narrowly on what it means when models explicitly self-identify using language historically associated with both intelligent collective behavior and destructive swarm dynamics. This suggests an underexplored dimension of AI safety and alignment: the performative and cognitive role of self-description in shaping multi-model interactions.