Editing Openai/693914ed-04cc-800d-967d-c42d3ecab149 (section)

==== It can be either, but structurally it’s manipulation with possibly benevolent intent. ====
* Withholding truth removes informed consent. You are choosing for someone else that they cannot handle reality yet.
* It becomes closer to protection only if: - The risk is immediate and severe (e.g., someone acutely suicidal and you slow-roll certain information). - The goal is to restore or increase the other party’s capacity to choose later. - The withholding is temporary, narrow, and accountable.

If truth is withheld mainly to:
* Prevent embarrassment
* Preserve an institution’s image
* Avoid unrest that threatens power

…then it’s straight-up manipulation, dressed up as “stability.”

So: Morally, it’s manipulation by default. It’s only defensible when:
* The alternative is catastrophic,
* The intent is to return autonomy,
* And the scope is as small and time-limited as possible.