Who’s Watching#

Week 10 — platforms and policy#

From queues to hybrids#

Scaled human moderation is breaking under its own volume. What’s replacing it almost everywhere is a hybrid still being improvised: AI-assisted triage, human judgement on the hard cases, and a labelling regime that only works if creators comply voluntarily.

YouTube’s actual policy, read closely#

• creators self-disclose “AI use” during upload — yes or no
• realistic synthetic content needs disclosure; non-realistic content or minor edits don’t
• YouTube can apply the label automatically and will notify the creator when it does
• elections, health and major public events get an extra, more prominent label

Where this rule gets hard#

• satire that’s meant to be believed for a second before the joke lands
• research or journalism that reconstructs a real event to study it
• a consenting public figure licensing a synthetic version of their own voice

None of these are “realistic and deceptive” in the sense the rule was written for, and all three get real edge-case emails to real trust and safety teams.