agent definitions — law

overview

legal treatments of agency define relationships between principals and agents, specifying authority, liability, and fiduciary duties. the cluster centers on restatements, statutes, and professional rules that constrain how humans (and potentially software) act on behalf of others.

signature traits

• delegated authority: agency Requires assent, control, and action on behalf of the principal, as codified in the restatement (second) and (third) of agency.
• fiduciary duties: loyalty and care obligations shape the agent’s autonomy and liability boundaries.
• institutional context: legal agency is embedded in contracts, corporate structures, and regulatory oversight, emphasizing enforceable accountability.

illustrative definitions

• 1958 — restatement (second) of agency: canonical black-letter definition, anchoring duty and control.
• 2006 — restatement (third) of agency (referenced by later entries): updates scope to modern organizational forms and clarifies principal control.
• legal ethics & regulatory sources: professional responsibility rules (e.g., rule 5.4) implicitly define acceptable agency relationships within law firms.

relation to other dimensions

• autonomy spectrum: legal definitions cluster on the low-to-moderate end; agents act with discretion but within clearly bounded mandates.
• entity frames: the principal–agent pair is a hybrid entity; human agents operate inside institutional scaffolding.
• goal dynamics: goals are typically supplied by principals; negotiation occurs through contracts rather than in-the-moment deliberation.
• persistence & embodiment: agency persists as long as authority is granted—whether the agent is a human employee, corporate officer, or potentially a software agent with delegated authority.

open questions

• how should fiduciary duties translate if software agents act autonomously on behalf of clients?
• do existing doctrines treat llm-based agents as instrumental tools or as entities capable of breaching duty?
• what regulatory updates are needed to signal when hybrid human–ai agency relationships are valid or void?