Trust Center / AI register

Every AI system on the platform, listed.

What each one does, whether it is switched on, how the EU AI Act treats it and who checks its decisions. Member data is never used to train AI models, by us or by any supplier. No generative AI is in use today.

Reviewed 18 October 2026

AI register

5 systems, 2 in use today.

In place: Operating today.
Not in use: Switched off for every partner until the date shown.

Automated moderation (Sightengine)

In place

What it does: Screens photos, videos and text for scams and prohibited content
AI Act: We deploy it. Not on the high-risk list in Annex III. Decisions that restrict a member can be appealed to a person (GDPR Article 22).
Who checks it: Our trust and safety team reviews every reported profile within 4 hours, and appeals (for every partner by January 2027)

Scam and fake profile detection

In place

What it does: Spots patterns linked to scams, fraud and fake accounts
AI Act: Not on the high-risk list in Annex III. A person decides on every reported profile.
Who checks it: Our trust and safety team

Facial age estimation (Yoti)

Not in use · until 30 Nov 2026

What it does: Will estimate a new member’s age from a selfie at sign-up
AI Act: Yoti is the provider and we deploy it. Members will be told before the check (Article 50(3)). How Annex III point 1(b) applies is being reviewed with our lawyers, ahead of the high-risk rules applying from 2 December 2027.
Who checks it: An ID document check when the estimate is close to the age limit

Face matching (Yoti)

Not in use · until 30 Nov 2026

What it does: Will confirm that a member’s face matches the photo on their ID
AI Act: One-to-one biometric verification, which Annex III point 1(a) leaves out of the high-risk list. Explicit consent under GDPR Article 9(2)(a).
Who checks it: Only with the member’s written consent

AI-written openers (None chosen)

Not in use

What it does: Would draft a first message from the other member’s profile, labelled “Written with AI”
AI Act: Would be labelled for members and marked in machine-readable form (Article 50). Not high-risk.
Who checks it: Off unless a partner switches it on; the member chooses whether to send. The model provider joins the sub-processor register, with 30 days’ notice, first

“In place” here means the system is in use today.

Our rules for AI

People stay in charge.

No banned practices. Nothing on the platform uses the practices Article 5 forbids, such as emotion recognition or manipulative techniques.
A person reviews every reported profile, and any automated decision can be appealed to a person (GDPR Article 22).
Labelled. Anything AI writes for a member is labelled for them and marked in machine-readable form (Article 50).
Never trained on member data, with written terms stopping AI providers from doing so (November 2026).
AI literacy. Staff who run or oversee these systems are trained on them (Article 4), as part of our security and privacy training from November 2026.
Nothing new without notice. A new AI system is added to this register, and its provider to the sub-processor register, before it is switched on.

The law

The EU AI Act, in short.

AI Act (EU) 2024/1689

In progress · Nov 2026

Official text: EUR-Lex

What matters: AI literacy (Article 4), banned practices (5), transparency (50, from 2 August 2026) and the high-risk list in Annex III (from 2 December 2027 after the 2026 amendments).
Your part: Tell members about AI in your own notices.
Our part: Keep the AI register current, label AI content, keep people in the loop.