Work

When Authentication Is Not Enough: Institutional Trust, Emergency Authority, and the Governance of Cybersecurity

Year: 2026
Type: article
Abstract: Cybersecurity extensively models authentication, authorization, integrity, provenance, and technical trust. It less often asks what happens when those mechanisms succeed but a human recipient still must decide whether an authenticated and authorized instruction warrants reliance. This paper develops that problem through a healthcare threat model. A legitimate security warning may be correctly authenticated, correctly delivered, and issued by an authorized security team, yet still fail operationally. The usual interpretation would locate the failure in the recipient: the employee did not trust legitimate authority. This paper argues that such an interpretation is incomplete. Rejection may be justified by institutional history, conflicting obligations, insufficient evidence, or previous organizational failures. The security-relevant object is therefore not compliance or institutional trust alone, but the governance of reliance under conditions of incomplete warrant. The problem becomes most acute during high-tempo incidents. When action is required faster than recipients can independently evaluate the evidentiary basis for an instruction, contemporaneous verification may become insufficient. Recipients may then rely partly on institutional history and on a credible expectation that extraordinary authority exercised under evidentiary constraint will later be made accountable. This paper terms that relationship trust credit: provisional reliance is extended under temporal constraint against an institutional obligation of later accounting. That solution creates a further problem. Who determines whether the original instruction was warranted? Who decides whether rejection was justified? Who determines whether evidence remains too sensitive to disclose? Who decides whether an action was sufficiently exceptional to trigger enhanced review? At each layer, the problem of legitimacy reappears as a problem of self-judgment. The paper therefore advances nested independence as its central governance principle. Nested independence does not require an infinite hierarchy of supposedly neutral reviewers. It requires non-concentration across seven governance functions: authority, review trigger, evaluative criteria, evidentiary control, assessment, disclosure control, and contestability. The architectural question is whether one actor controls enough of these functions to make its own exercise of authority effectively self-validating. Contemporary language models provide the motivating case rather than the governing subject. Model evaluations that operationalize cultural or political values expose the same underlying problem: whoever defines the evaluator can substantially shape what the evaluation recognizes as desirable, propagandistic, trustworthy, or legitimate. The paper therefore connects AI evaluation governance, healthcare cybersecurity, institutional trust, and emergency authority through a common question: Who controls the path by which an exercise of authority becomes recognized as justified? The defensive objective is neither maximum trust nor maximum institutional obedience. It is governable reliance: security arrangements in which extraordinary authority can operate when genuinely necessary without allowing the exercising authority unilateral control over the later determination of whether that reliance was warranted.
Language: English
Open Access status: green
OpenAlex ID: W7235533141
FWCI:
Cites:
Cited by:
Related to: 10