High-risk AI-mediated decisions often fail not because models are opaque, but because decision events are recorded as telemetry rather than as contestable semantic handovers. This working paper specifies decision-event transparency requirements and an operational “contestable record” framing for AI-mediated knowledge work. This Zenodo record archives the manuscript as a citable public artifact for independent verification. Journal outcome note: the manuscript was submitted to AI & SOCIETY and subsequently rejected; this deposit is not a claim of journal publication or endorsement. The aim is to provide a stable PublicLife anchor that can be audited and referenced independently of venue status. How to cite So, H. B. (2026). From Telemetry to Contestable Records: Decision-Event Transparency Requirements for AI-Mediated Knowledge Work (v1.5). Zenodo. https://doi.org/10.5281/zenodo.18602449 Public Works List So, Hon Bor — “From Telemetry to Contestable Records: Decision-Event Transparency Requirements for AI-Mediated Knowledge Work” (Working paper; Zenodo deposit; v1.5; public anchor). DOI: 10.5281/zenodo.18602449. Note: previously submitted to AI & SOCIETY; outcome = rejected (venue status independent of this deposit). E6/SMP SMP-UPDATE: Added Zenodo public working-paper anchor for “From Telemetry to Contestable Records: Decision-Event Transparency Requirements for AI-Mediated Knowledge Work” — DOI 10.5281/zenodo.18602449 (v1.5). Scope: public verification anchor; no journal-publication claim; prior AI & SOCIETY outcome recorded as rejected. Attachments: deposited manuscript files as per Zenodo record. (metadata-only)