Personal data leaves your control
Names, emails, NHS and National Insurance numbers end up in a third party’s systems, often outside your data protection agreements.
Tidewall checks every message the moment you press Send in ChatGPT, Claude, Copilot and other AI tools. Customer details and secrets are caught and redacted in one click, and nothing you type ever leaves your browser.
Draft a reply to [PERSON_1] ([EMAIL_1]) about her claim. Her NHS number is [NHS_NUMBER_1]. Use key [REMOVED].
Not out of carelessness. AI tools are useful, and the quickest way to get help with a letter, a spreadsheet or a bug is to paste the real thing.
Names, emails, NHS and National Insurance numbers end up in a third party’s systems, often outside your data protection agreements.
API keys, passwords and database URLs travel with the snippet someone wanted debugged.
Blocked at work, people use their phones, where you can’t see or guide anything. Tidewall lets them keep using AI, safely.
No new app to learn. Tidewall waits for the moment you press Send.
Ask ChatGPT, Claude or Copilot whatever you need. Tidewall stays out of the way.
When you press Send it checks the message on your device and shows what it found, and exactly what the AI would see.
Details become placeholders like [EMAIL_1]. In the reply, Tidewall puts the real details back on your screen only.
This box runs Tidewall’s real detection engine in your browser. What you type is not sent anywhere; you can check with your browser’s network tools.
Numbers are checked with their real checksums, so an order number isn’t mistaken for an NHS number.
Names, email addresses, phone numbers, postcodes, National Insurance numbers and IP addresses.
NHS numbers, payment cards, IBANs, sort codes and account numbers, each verified before it’s flagged.
API keys and tokens for 20+ services (cloud, AI, payments, code), private keys, passwords and database URLs.
Project names, client codes and patterns like ACME-CL-#####, set by you or your IT team.
Allow, Warn, Redact or Block for each kind of data. Warnings can ask for a reason before sending anyway.
Mark tools as approved or blocked. A blocked tool shows a friendly page pointing to the approved ones.
A tool that stops data leaking shouldn’t collect your data. So Tidewall doesn’t.
Push Tidewall with Google Workspace or Microsoft Intune and set the rules once. Staff can see them but not switch them off.
{ "orgName": "Acme Health", "actions": { "email": "redact", "nhs": "redact", "card": "block", "secret": "block" }, "customWords": ["Project Neptune", "ACME-CL-#####"], "sites": { "character.ai": "blocked" }, "requireReason": true }
Free for individuals. A few pounds a person for teams. Prices exclude VAT.
Founding customers: half price for your first year, for the first 25 organisations, in return for a short case study.
No. The checks run inside your browser. Tidewall never stores or sends your messages, and the extension makes no network requests of its own.
ChatGPT, Claude, Gemini, Google AI Studio, NotebookLM, Microsoft Copilot, Perplexity, DeepSeek, Mistral Le Chat, Grok, Meta AI, Poe, Hugging Face Chat and more. We add tools as they become popular.
Checks take about a millisecond. Messages with nothing sensitive go straight through; you only see Tidewall when it has found something.
No tool can. Tidewall verifies numbers with checksums, recognises names from many cultures and context, and lets you add your own client codes and project names. It turns “we hope people are careful” into a control you can show an auditor.
On the Team plan, rules are set by your organisation through your browser management and can’t be paused or changed by staff.
Tidewall helps you meet your obligations by stopping personal data reaching AI tools you don’t have agreements with. Because it doesn’t collect personal data itself, it doesn’t add a new processor to your records.
Not yet. Tidewall checks text you type or paste. File and image checks are on the roadmap.
Free for individuals. Set up for a whole team in minutes.