AI at SHY

Our AI principles

SHY uses AI where it helps people understand each other — and deliberately keeps it out of decisions about consent, safety, and access.

1. AI never decides consent or safety

Who can message you, who sees you, blocks, partner consent for couples, and enforcement decisions are made by deterministic server code — rules you can read about on this site — never by a model's judgment.

2. No invented numbers

When SHY doesn't have a real signal about compatibility, it says nothing rather than fabricating a score. There is no public "match percentage" on people; Harmony explains grounded reasons — like values you both actually selected — or stays quiet.

3. Authenticity over generation

Using AI to misrepresent yourself — generated photos presented as you, impersonation, or automated conversations pretending to be human — violates our Community Guidelines.

4. Consent-gated assistance

AI assistance in SHY (such as conversation suggestions or plan ideas) is opt-in, and the assistant only reads what you've consented for it to read. Your private message content is not used to train models.

5. Review before change

Changes to matching behavior go through review for unintended bias across orientation and demographic groups before shipping, and every feed carries an internal ranking-policy version so changes are attributable and reversible.

Last reviewed: August 19, 2026.