AI transparency

When you talk with HUMANS, you are talking with an AI agent, not a human. It is designed to get to know you through conversation and build a private profile that helps us find who fits you. It has limits: it can misunderstand, and it is not a therapist, a doctor, or a crisis service.

Safety-relevant conversations may be reviewed by a person on our team. If the AI detects signs of crisis, it responds with care and shows you crisis resources. See our crisis protocol.

Your conversation and profile text are processed by the AI providers listed below to run your conversation, build your profile, and evaluate compatibility. Your photos are never sent to any AI provider. Matching evaluations carry first names only, never contact details.

• OpenAI: ai conversation + profile extraction (conversation text, profile text). API data is not used to train models (OpenAI API policy).
• Anthropic: alternate ai provider (conversation text, profile text). Alternate to OpenAI; API data is not used for training under its Commercial Terms.
• Voyage AI: search indexing (embeddings) (profile text). Training opt-out enabled on our account (effective 2026-07-19); data is deleted after processing.

The full list of everyone who processes your data is on our subprocessors page.

HUMANS builds a private psychological profile from your conversation. An automated system applies hard filters (age range, distance, and the orientation you are looking for) and then scores compatibility to decide which proposals you see. Photos and contact details are never part of the evaluation.

Information the system uses:

• Your structured profile, built from your conversation with the AI
• A narrative summary of who you are and what you are looking for
• Your stated dealbreakers
• Hard filters you set: age range, distance, and the orientation you are looking for
• Your first name, only so the written explanation can refer to each of you naturally

Information it does not use:

• Your photos (the matching AI never sees them)
• Your contact details
• Protected attributes as a basis for exclusion, beyond the dealbreakers you set yourself

Every proposal comes with a written explanation from the matching AI. You have the right to ask how a decision was made and to have a person review it. In the app, use “request a human review” on a proposal or in Settings. Without an account, write to us through our contact page. A person replies within 15 days.

The compatibility explanations you read are generated by AI. We label this so it is always clear which text a machine wrote.