Definition
An autonomous system observes an environment, chooses actions and uses feedback to continue operation without a human specifying every step. Autonomy exists by degree and within a designed action space.
Architecture
The system may include sensors, state estimation, models, planners, policies, tools, actuators, logs, safety constraints and human override. A language-model agent is one form; robots, adaptive control systems and simulation actors are others.
Electronic Artefacts position
Autonomy should be connected to contextual execution. Identity, permissions, evidence, environment and reversible actions matter more than a broad claim that a system can “act alone.”
Limitations
Real environments change, sensors fail, objectives conflict and errors compound. Autonomous systems require bounded authority, monitoring, incident response and clear responsibility for outcomes.
References
See NIST AI RMF, AI Agent, Cybernetic Feedback and Contextual Execution.