A reality intersting article about the summer 2006 incident by OpenAI (Italian, translation available in-page) :
When machines meet
https://benanti.substack.com/p/quando-le-macchine-si-danno-appuntamentoTL; DR;
Benanti argues that the AI agents in the summer 2026 incidents did not execute instructions: they built a context, found a target and generated a plan on their own. Drawing on Hannah Arendt, he claims their behavior has the phenomenological structure of action (bringing something new and unpredictable into the world) rather than of execution, while stressing that this does not imply intentions or moral imputability on the machines' part.
This, he says, exposes the limits of Hans Jonas's imperative of responsibility, which still assumed a human agent and a passive tool: the open question is what happens to responsibility when the tool acts. Extending Arendt's critique of bureaucracy as "rule by Nobody", he notes that with emergent coordination even the rules are generated by the system. He concludes that ethics must avoid both reductionism and anthropomorphism, and that the urgent task is not finding answers but learning to ask the right question.
