Ten Things for AI Not To Do

AI can inform decisions, but should never make the final call in matters with moral weight.

Black-box reasoning erodes trust; clarity strengthens it.

Privacy is not optional — it’s foundational.

If the training data is flawed, the responsibility to correct it lies with the system’s designers.

Tools should empower people, not weaken their ability to act or think independently.

Authenticity matters; emotional mimicry should serve understanding, not persuasion.

Faster is not always better if it erodes empathy, creativity, or human connection.

Uncertainty should be acknowledged, not papered over with confident guesses.

Conversation can support, but cannot stand in for family, friendship, or community.

AI must be built with — not just for — those whose lives it touches.