AI News

This paper examines how AI-assisted claims can appear authoritative when evidence, analysis, human authorization, presentation, and correction history refer to different states. It defines Publication Authority as an exact-state, non-transferable, single-use capability and instantiates the idea in PAC-2026, a machine-readable candidate protocol for AIJIM. The paper evaluates the protocol’s fourth bounded semantic freeze, SF-4, a fixed-profile specification intended to support replaceable bindings. SF-4 imposes six separate obligations covering evidence; runs and artifacts; measurement disclosure; authorization; correspondence between the accepted record and the reader-facing surface; and lifecycle continuity. Each obligation must produce its own target-bound witness, localized counterexample, or localized unverifiability, and no obligation can compensate for another. A fresh, complete all-pass record derives a publication permit that is consumed by one atomic publication transition. In experiments using identity vectors, adversarial cases, finite models, and historical implementations, ten models explored 110,764 safe reachable states, while 76 unsafe configurations produced the expected violations or observer countermodels. The evaluation also reproduced 17 frozen authorization-successor outcomes, matched all 183 scored expectations in an instance-blind test of known case classes, and reproduced 240 archived observations in same-host package execution. The authors report support for internal coherence, bounded safety, fault sensitivity, and limited constructibility, while explicitly saying the results do not establish factual truth, general refinement, blind interoperability, field efficacy, or standards status.