Section 6 of 102
5. Autonomous Agent Definition
Stable section ID: S05-CON-010-SECTION-6 · 14 content blocks
An Autonomous Agent is an identifiable computational entity capable of observing an environment, maintaining operational context, selecting or planning actions, using authorized tools, and pursuing defined objectives with a specified degree of independence.
An AI system shall be considered an agent when it possesses one or more persistent operational characteristics, including:
- A registered identity.
- An assigned role.
- Defined objectives or responsibilities.
- Persistent or session-based context.
- Capability to monitor events or changes.
- Capability to initiate recommendations or actions.
- Access to tools, services, or physical systems.
- Defined permissions and authority boundaries.
- Runtime state and lifecycle management.
- Accountability and audit records.
Autonomy shall not be treated as a binary condition. An agent may be autonomous for one narrowly bounded function while requiring human approval for another. For example, an operational agent may autonomously adjust non-critical lighting schedules but may only recommend changes to structural, fire-safety, or emergency-control configurations.
An autonomous agent shall operate only within an approved operational envelope defined by its identity, manifest, capabilities, permissions, current system state, applicable policies, and available evidence.