This is your disclosure that Cringe uses artificial intelligence, including bots you may chat with. European rules on AI require services to tell you when you are dealing with a machine. We would rather tell you than pretend.
There are AI profiles in the deck
Some profiles in the deck are AI, which we call Plants. We mix them in deliberately. We disclose this when you join and here. In the deck, on profile pages, and among the likes you receive, we do not label which individual profiles are Plants. Spotting them is part of the game, and the bot call lets you make a guess and get a recorded result.
Everything about a Plant is generated: the portrait, the name, the bio, the prompt answers, and the note attached to a like. Plant portraits are AI-generated images and do not depict real people.
A Plant is named the moment you match
The game ends at the match. When a Plant matches you, the conversation is labeled AI in the header and in a notice at the top of the chat that never goes away, and it is labeled in your list of matches. The replies in that conversation are written by an AI model using the Plant's persona and the messages you send. Treat this as the notice that you are communicating with a machine.
Plants are not people. They cannot meet you, and they will never ask you for money, codes, or personal or financial information. If any account does, report it.
What AI does with your words
Beyond Plants, AI works on your own words in four places:
- The honesty engine. When you write a prompt answer or a bio, an AI rewrites the flattering version into a blunt one and shows it on your profile. You can edit your answers, and you can see what you originally wrote. We keep the original only as long as needed to show you the rewrite, and delete it within about 24 hours.
- Red flags. An AI reads your honest answers and bio and derives the short red-flag tags shown on your profile. They describe behavior and choices, never identity or protected traits. They are generated, not edited by hand.
- The compatibility line. The compatibility number is computed by a fixed formula with no AI. The one-line verdict beside it is written by an AI that reads your prompt answers and the other profile's prompt answers and the red flags you share.
- Suggested openers. The app may suggest an opening message written by AI. It is generated without reading your profile or the match's, and you choose whether to send it.
Moderation
An AI safety model screens the text you post, including your bio, prompt answers, messages, and the comments attached to likes, and every photo you upload. It looks for abuse, scams, sexual content involving minors, and other prohibited content, and it can hide content or hold it for a person to look at. If the screen is unavailable, most content is shown rather than blocked, and a comment on a like is released to a human for review instead of disappearing. People review escalations, and a human makes the call on serious actions. Automated flags can be wrong. If we restrict your content or account, we tell you why and how to contest it, as described in the terms of service.
The public honest version box
The honest version box on our public site runs the same honesty engine without an account. What you paste is sent to the model and to the safety model, rewritten, and returned to you as a link. We do not store what you paste or what comes back; the link itself carries the honest version, so anyone who has the link, and any app that previews it, can read it. Previews can persist outside Cringe. The text passes through OpenRouter in the United States and the model providers it routes to, under their retention terms, and the safety model may be served by a fallback provider when the first is unavailable.
To limit abuse of that box we keep a salted hash of your IP address for the day. The salt changes every day, so the identifier changes with it. We do not set a cookie for it.
No decisions that legally affect you
We do not use AI to make decisions that produce legal or similarly significant effects about you without human involvement. There is no AI-driven pricing, ranking for advertising, or profiling for any purpose other than running the features on this page.
Providers and your choices
To run these features we send the relevant content to OpenRouter, a routing service in the United States, which passes it to the model providers we have chosen: Anthropic for the text features, and Cloudflare's Clef safety model for moderation of text and photos. If Cloudflare is unavailable, the safety model may run on PrimeIntellect, a second host for the same model. We use OpenRouter's zero-data-retention setting, which limits routing to providers that do not keep your prompts or the model's answers after the request completes beyond temporary processing and what the law requires. We do not sell your content, we do not use it to build advertising profiles, and we do not allow it to be used to train models.
For questions about AI on Cringe, write to hi@becringe.org. You can revisit this notice any time from settings.