Legal
AI Disclosures
Last updated: October 15, 2026
Job Hunt OS includes assistive features that suggest, extract, and draft. This page explains each one conservatively. Every output is a suggestion for you to review; none is guaranteed to be correct, and none makes decisions about you.
1. Who processes your content
When AI assistance is on (the default), the text needed for the feature you start — for example, text or a résumé you submit for fact extraction, a job description you paste, or your confirmed facts and a résumé version when you ask for suggestions — is sent to Anthropic, PBC (Claude) for processing. Under Anthropic's commercial API terms, your content is not used to train its models; Anthropic may keep it for a limited period for safety and abuse monitoring (XXX days, pending confirmation). Your work-authorization note, your contacts' details, and your sign-in data are never sent. You can turn AI assistance off in Settings; the built-in, rule-based assistant is then used and nothing is sent to Anthropic.
Anthropic, PBC's model is used only for the three features marked “Claude” below. Everything else runs on Job Hunt OS's built-in, deterministic rules on our servers. If a Claude request fails or AI assistance is off, the rule-based assistant is used instead.
2. Where assistance is used
| Feature | What it does | What it processes | Runs on |
|---|---|---|---|
| Fact extraction | Suggests career facts from text you write or a résumé you upload. Each suggestion quotes your words and starts as unconfirmed. | The text or document you submit | Claude |
| LinkedIn import | Reads the LinkedIn profile PDF or data-export files you upload with fixed rules and turns each entry into a suggested fact that quotes the file. Fills only empty profile fields (name, headline, summary, city). | The LinkedIn files you upload | Built-in rules |
| Job-description parsing | Splits a job description into requirements (required vs. preferred, years, categories). | The job description you paste | Claude |
| Profile alignment | Compares requirements with your confirmed facts and shows which are met, partly met, missing, or unknown, with a weighted, visible formula. | Requirements and your confirmed facts | Built-in rules |
| Résumé suggestions and tailoring | Suggests reordering, tightening, or rephrasing bullets for a role, using only facts you confirmed. You accept, edit, or reject each suggestion. | Your résumé version, confirmed facts, and the job requirements | Claude |
| Template recommendation | Recommends a template family based on the role's function and seniority. | Role target information | Built-in rules |
| Outreach and thank-you drafts | Fills a fixed template with confirmed facts. Contact names are inserted locally and never sent to Anthropic. | Your confirmed facts and the opportunity | Built-in rules |
| Interview preparation | Lists likely questions from the role's requirements and lets you pick stories you wrote. Improvement themes are suggested from your own reflection text and saved only if you confirm them. | Requirements, your stories and reflections | Built-in rules |
3. Limits of automated suggestions
- Outputs can be incomplete, outdated, or wrong. Review everything before you use it.
- Suggestions are informational. They are not career, legal, HR, immigration, tax, or financial advice.
- The alignment view is not an applicant-tracking (“ATS”) score and does not predict whether you will get an interview or offer.
- No feature analyzes emotions, personality, or protected characteristics, or makes inferences about them.
- Nothing is sent, submitted, or published on your behalf.
4. Safeguards built into the product
- Extracted information is never used until you confirm it.
- Suggestions that add a claim must cite facts you confirmed; suggestions that would add new numbers, dates, employers, or skills are discarded.
- Instructions hidden inside documents or job descriptions are treated as plain text and flagged, not followed.
- Résumé layout is produced by fixed templates; automated features never generate code or formatting.
- We record which feature ran, when, and its status and structured result to operate and debug the service — not to profile you.
5. Your choices
You can turn AI assistance off in Settings → Preferences; the app then uses only the built-in rules. You can delete any suggestion, draft, or extracted fact. Questions: XXX (privacy email pending confirmation).