Responsible AI

Kyrolane can rank jobs against your profile, parse a resume you upload, and draft a cover letter or a tailored resume when you ask. In the Partner Portal, recruiters can rank candidates against a role and draft an outreach note.

These tools assist. A person decides whether to send, whether to interview, and whether to hire. We do not make a solely automated decision that produces a legal or similarly significant effect about you.

A draft can be wrong or out of date. Read it before you send it or rely on it. When you generate one, we send the resume and job text needed for that request to the configured model provider. With the platform key, that provider is OpenAI. We do not train a Kyrolane model on your mailbox.

Chat and draft requests set store to false where the OpenAI chat API supports that parameter, so the provider is asked not to keep the output for distillation or evals. We do not send a setting that opts the request into model training. Embedding requests do not have a store parameter.

The default partner score is a keyword match. The product labels it that way while semantic matching is off.

Semantic matching compares meaning using OpenAI embeddings (text-embedding-3-small) and blends them with the keyword score. It runs only when the server flag is the exact string true. A key being present does not turn it on. The default blend weight is 0.35, and there is a daily cap on embedding calls.

Talent-pool matching includes a job seeker only after they opt in. New accounts are off. People who never stored a preference are not eligible.

• Job seekers approve each application before it leaves their mailbox. Follow-ups stay off until enabled.
• Partner Recruiter Assistant will not send until a recruiter approves the message. The footer says the message was drafted with automated matching and a recruiter chose to send it.
• Partner follow-ups, if enabled, send a template a person wrote, not a new generated draft.
• Job match results and partner candidate match results include a Request human review button. It saves a review request and emails privacy@kyrolane.com. We then look at the score inputs we still have.

If the law treats your use of ranking as an automated employment decision tool, or as a system with deployer duties, those duties are yours. Examples that may apply, depending on your facts, include New York City Local Law 144 and the EU AI Act. We do not file bias audits or candidate notices for you. See the Partner Terms.

Scores reflect the text we have. They can miss context, prefer common phrasing, and lag behind a market. Do not treat a score as a measure of a person’s worth or of a guaranteed fit.