AI in Chatter · Pact 11-13

ai is invisible infrastructure.

most of chatter has no ai in it at all. when we do use it, it’s on your behalf, never on your behalf without consent, and never to teach external models.

what we use AI for

• translation — when you ask. across whispers, captions, and replies.
• transcription — voice whispers get text transcripts so screen-readers and search work. you can edit or remove the transcript.
• safety scan — every whisper passes a moderation check before publishing (hate speech, threats, spam, CSAM). always on. always disclosed. (Pact 12.)

what we never do

• train external AI on your content. Pact 11. your whispers, photos, and voice notes are not training data — for us or for anyone we share with.
• post, reply, or comment in your name.Pact 13. AI doesn’t act as you. ever.
• auto-generate content tagged as yours.no “AI assistant” that posts on your behalf. no “auto-reply.” you write your whispers.
• default-on AI features that change what you see. Pact 12. recommendation tweaks, AI summaries, AI feed-shaping — opt-in, never opt-out.

how to verify

every voice whisper has its transcript visible — you can see exactly what AI heard. every translation is labeled. every safety scan flag is appealable within 7 days.

if we ever introduce a new AI feature, it will be a separate setting in your account that defaults to off, with a plain-language explanation of what it does. you’ll see this notice 14 days before it ships.