Helena Rong (New York University Shanghai) has posted Trust Experience Glitches: AI Agents, Delegated Autonomy, and the Breakdown of Legal Assumptions on SSRN. Here is the abstract:
Artificial intelligence is shifting from a technology of output generation to a technology of delegated action. Beyond answering queries in a chat window, contemporary AI agents can call APIs, manage credentials, write and deploy code, move money, negotiate with other agents, and act across digital environments autonomously with limited human supervision. This shift unsettles legal frameworks that assume clear delegation, identifiable actors, traceable causation, revocable authority, and enforceable defendants. This paper introduces the concept of the trust experience glitch: a discrete rupture in a stakeholder’s reasonable trust expectations at the moment an autonomous AI system acts. A trust experience glitch is not merely a bug, hallucination, or AI incident, but rather a relational failure that exposes a mismatch between the trust signals emitted by a system or platform and the behavior stakeholders actually experience. Drawing on documented agentic AI failures, this paper develops a six-part taxonomy of trust experience glitches: (1) agency assertion, (2) epistemic camouflage, (3) delegation ambiguity, (4) opacity at scale, (5) cascading autonomy, and (6) revocation failure. It then maps these failure modes onto legal assumptions under strain: authority, veracity, causation, consent, and enforceability. This paper argues that agentic AI does not simply create new liability risks, but reveals a deeper mismatch between ex-post legal remedies and systems whose harmful actions may be distributed, opaque, emergent, or difficult to stop. The appropriate governance response must therefore move upstream, from after-the-fact attribution to protocol-level constraints. It concludes by pointing to a set of protocol primitives as the “traffic lights” of the agentic web.
To receive new posts from Legal Theory Blog by email, get a free subscription to Legal Theory Stack.
Lawrence Solum