Veyl — Mask before you ask

Masks sensitive data in ChatGPT and Claude before you send. Real values restored locally. Nothing leaves your browser unmasked.

Every professional knows the moment: AI could draft this memo in seconds, but the text contains a client's name, SSN, or account number — information you're bound to protect. Veyl ends that trade-off. It works quietly inside AI models like ChatGPT and Claude — As you type, Veyl detects sensitive information: SSNs, EINs, ITINs, payment cards, account numbers, medical record numbers, dates of birth, phones, emails, addresses, policy numbers, tax preparer IDs — and names, using an AI model that runs entirely on your device. — When you hit send, Veyl asks: "Sensitive information detected — mask before you ask?" Review each item, unmask what you choose, mask anything it missed with one click. — Only the masked version leaves. The AI sees [PERSON_1], never the name. — The AI's answer is restored on your screen, locally. The real values never left your computer. — Every protected send is recorded in a tamper-evident, hash-chained audit log: categories and fingerprints, never your text. PRIVACY, ARCHITECTURALLY Veyl has no server that sees your text. Detection, masking, restoration, and the audit log all run in your browser. The only network activity is a one-time download of the local AI model. We cannot read your data — not as a promise, but as a design. FREE FOREVER Pattern detection, AI name detection, mask/unmask control, local restoration. PRO (COMING SOON) Document masking (PDF/Word), custom always-mask dictionary, variant matching, audit report export. Built for professionals bound by confidentiality like CPAs, attorneys, valuation and research professionals and for anyone who wants what they type into AI to stay theirs. Veyl is an early-stage product. Use judgment with highly sensitive production data. www.projectveyl.com