Retained Authority
Retained authority is Diplomats.Digital's institutional test for whether an organization that uses AI or digital systems remains able to define objectives, set boundaries, authenticate actors, approve consequential action, verify outputs, intervene or override, trace decisions, maintain continuity, and exit dependencies. Authority is retained only when these powers remain operationally executable, not merely stated in policy or contract.
An institution can remain legally responsible while losing practical control over the systems through which decisions and operations are carried out. Retained authority examines that gap. It asks who can act, on what basis, through which system, with what evidence, under whose approval, and what happens when a tool, vendor, model, or communication channel fails or changes.
Foreign-affairs institutions cannot delegate accountability to a model or vendor. Diplomatic positions, crisis decisions, official communications, sensitive analysis, citizen services, and intergovernmental commitments require clear institutional ownership. Retained authority provides a more precise test than general claims of sovereignty or responsible AI because it focuses on the mechanisms through which authority is exercised and preserved.
Retained authority appears in use-case approval, role and identity controls, human authorization points, AI and system logs, provenance records, escalation protocols, vendor contracts, data portability, continuity arrangements, override mechanisms, fallback procedures, and the institution's capacity to replace or discontinue a system without losing essential functions or records.
Closely related entries kept separate because each carries a distinct institutional meaning.
- Human Oversight
Human oversight refers to the meaningful ability of people to understand, review, question, approve, intervene in, or stop the use of automated or AI-supported systems. In foreign affairs, it ensures that technology supports diplomatic judgment rather than replacing responsibility.
- AI Sovereignty
AI sovereignty is the capacity of a state or institution to make meaningful, informed, and accountable choices about the models, data, compute, infrastructure, vendors, deployment arrangements, and exit options on which its use of AI depends. It does not require technological self-sufficiency. It requires enough capability, leverage, and control to choose, govern, verify, change, or leave those arrangements while protecting strategic choice and public authority.
- Exit Capacity
Exit capacity is the institutional ability to leave or replace a system, technology, platform, AI model, or provider while preserving data, operational continuity, knowledge, security, legal compliance, and authority. An exit plan is documentary; exit capacity must be operational, adequately resourced, and testable.