Privacy by Design: What Happens to Your CV After You Upload It

Uploading a CV is an act of trust. You are handing over employment history, contact details, maybe salary hints, visa status, gaps you would rather not explain to strangers. Any CV data privacy story that starts with "do not worry" and ends without specifics deserves scepticism. This article explains what actually happens inside Cvaluate after you click upload — what we store, what we send to AI providers, what we do not do, and how that lines up with our privacy policy.

What you send us

A typical session includes: your PDF (or other supported file), an optional job description, and your Google sign-in identity so we can attach results to your account. The file may contain special-category data depending on what you wrote — health-related career breaks, union membership, diversity fellowship names. We treat the whole upload as sensitive résumé upload privacy data regardless.

We do not require payment details for the free tier. We do not ask for national ID numbers or references at upload. Minimise what you include on any CV; we analyse what you provide.

What happens during analysis

At a high level:

1. Ingest and store — the file is saved securely against your user ID so you can revisit results.
2. Extract text — parsing runs on our infrastructure (see PDF extraction challenges).
3. Analyse via AI API — structured extracts and prompts are sent to Anthropic's API for scoring and feedback generation.
4. Validate and persist results — JSON output is checked, then saved for display in your history.

We do not manually read your CV for routine quality control. Engineering access to production data is restricted and logged. Aggregated, anonymised telemetry (error rates, latency) does not include CV content.

Third-party AI processing

Cvaluate is not a self-hosted model on a laptop in a cupboard. We use commercial AI APIs — currently Anthropic Claude — under data processing terms appropriate for customer content. API providers contractually restrict using API inputs to train their public models. Your CV text leaves our servers only for the purpose of generating your analysis, over encrypted connections.

This is the same class of trade-off as using any cloud SaaS: you gain capability; you rely on vendor security and contractual safeguards. We document subprocessors and transfer mechanisms in our privacy policy for GDPR résumé tool users who need transparency.

What we do not do

• Sell or rent CVs to recruiters, data brokers, or advertisers
• Build a searchable talent pool from uploads
• Publish your results or make them discoverable on the open web
• Use your content to train a foundation model we ship to others
• Share data except as required by law or described in the privacy policy

Our business model is helping you improve applications, not monetising your file twice. If that ever changed, it would require explicit consent — not a buried policy update.

Retention and deletion

Data retention exists so you can compare scores after edits, re-open rewrites, and avoid re-uploading the same PDF fourteen times. You can delete individual analyses from history; you can delete your account. Backups roll off on a schedule described legally — we do not keep CVs forever "just because".

If you need a copy of your data or want everything removed, contact paths are in the privacy policy. We aim to honour deletion requests without dark patterns.

Security basics

Transport encryption (HTTPS), access controls on stored files, authentication via Google OAuth, and separation between public marketing pages and authenticated product routes. No system is perfectly secure; we fix vulnerabilities and limit blast radius when credentials leak. Do not reuse a CV that contains secrets you would not email to a stranger.

Privacy by design in product decisions

Privacy is not a legal page stapled on at launch. It shapes features: authenticated-only results pages excluded from search indexing; no public share links by default; caching scoped per user; prompts instructed not to echo unnecessary personal data in feedback. Speed optimisations — see speed vs depth — must not leak one user's extraction to another.

We also avoid dark UX: clear sign-in, clear history, clear delete. Trust compounds when engineering and legal tell the same story.

What you can do

• Read the full privacy policy before uploading sensitive details.
• Redact information you do not want processed (e.g. full address if unnecessary).
• Delete old analyses you no longer need.
• Use a CV tailored for applications — not a dossier of every job since 1998.

AI hiring tools are powerful and opaque by default. We try to be the opposite on data: specific claims, written down, reviewable. For how analysis works technically — without the privacy lens — start with our guide to AI in hiring and hallucination guardrails. When you are comfortable, run your CV through Cvaluate's free analysis — your document, your account, your control.