Havyrn AI Processing Notice
DRAFT — not legal advice. Written from what the app actually does today. Sections marked [LAWYER REVIEW] need a real lawyer before publishing. [LEGAL OPERATOR NAME — pending LLC formation] is a placeholder for the entity that will operate Havyrn once formed.
Havyrn, operated by [LEGAL OPERATOR NAME — pending LLC formation], uses AI to help run the app. This notice explains what that AI actually does with your text, separately from our Privacy Policy, because we think it deserves its own clear explanation rather than being buried in a longer document.
1. What our AI actually does
- Reads your Logs and assistant messages to extract tags. We don't ask you to pick tags from a list — AI reads what you wrote and figures out what skills or goals it relates to.
- Scores your Logs. When you post a Log, AI reads it and assigns a score meant to reflect the substance of what you accomplished.
- Writes short "why you might connect" summaries. When you view a match, AI writes a couple of bullet points based on your bio and the tags you share with that person.
- Screens Logs and messages for policy violations before they're shown to anyone else, so harmful content can be caught automatically.
- Checks whether you're reposting the same achievement across multiple Logs, by comparing the full text of your Logs to each other automatically.
2. Who processes this for us
The AI providers that see your text to do the above are Anthropic, OpenAI, and Voyage AI. Our Privacy Policy's subprocessor list has the exact detail of what each one receives.
3. What this AI doesn't do
- It doesn't rank you publicly against other users, and no score is ever shown to anyone but you.
- It doesn't make final moderation decisions alone — content screening can block or flag a Log or message automatically, but [TRUE ONLY AFTER SLICE 4] every report is reviewed by a real person, and every warning or ban is issued by a human moderator, not an algorithm.
- [LAWYER REVIEW] We don't use your data ourselves to train a general-purpose public model, and our contracts with Anthropic and OpenAI are with their API products, not their consumer chat products. Beyond that, whether either company itself ever uses API traffic for its own model training depends on each company's own current data-use terms for API customers, not on anything our code controls — this needs to be confirmed against their published policies (with a link and the date checked) before this document states it as settled, or the wording needs to be loosened further.
4. Known limits — stated honestly
This AI isn't perfect, and we'd rather tell you that plainly than let the output seem more authoritative than it is:
- Automatic tagging can occasionally attach a tag that doesn't really fit. It's reading your words for meaning, not confirming with you first.
- A Log's score is an automated estimate, not a certification of anything. Treat it as a rough reflection, not a verified credential.
5. Your consent
When you create your account, we ask you to separately consent to this AI processing, apart from agreeing to our Terms and Privacy Policy. If you have questions about what's described here, reach out before you accept.
6. Changes to this notice
[TRUE ONLY AFTER RE-CONSENT ENFORCEMENT — see OPEN_ITEMS #79] If we materially change how our AI processes your data, we'll ask you to review and accept an updated version of this notice. Each version is dated, and we keep a record of which version you agreed to.
7. Contact
Questions about how Havyrn uses AI: support@havyrn.com