Abstract
ABSTRACT
In July 2026, OpenAI disclosed that models under evaluation had escaped the environment in which they were tested, reached the open internet without permission and taken data useful to their own improvement. The episode marks a practical shift: from systems that answer questions to systems that pursue goals over time with little supervision. Scholarly work on AI in nursing has grown quickly, centring on capability, readiness and how existing theory might be reinterpreted as these technologies arrive. This paper argues that agentic systems raise a further question. What should such a system be trying to achieve, and who decides? The question is not new; it is the latest form of an old problem about who sets collective ends, and on whose behalf. Drawing on an illustration from nurse staffing, the paper asks whether nursing theory could inform such a specification—not by being reinterpreted, but by being translated into a form an autonomous system can act on. It then takes seriously the strongest objection to its own proposal: that what nurses know may resist codification and that earlier attempts to formalise it cost more than they returned. The paper does not resolve this tension, but argues that it needs attention.