Privacy-first AI · two products

Use any AI. Expose nothing.

Start free in your browser, or run the whole thing as infrastructure. Either way, every name, email and identifier is stripped before a single token leaves your walls.

Region: EU Retention: Zero Detection runs locally

Your team wants ChatGPT. Compliance says no.

Every useful prompt carries client names, contracts, patient records, card numbers. Paste that into a public model and it can be logged, trained on, or subpoenaed. So the safest teams just… don't — and lose the productivity to the firms that took the risk. Both products below remove the reason to say no. One takes thirty seconds; the other takes an afternoon.

Live redaction — try it
your prompt
0 identifiers found
What the model sees
Edit the text, hit SCAN — detection runs in your browser. Nothing is sent.

Free · Chrome Web Store · v2.1.0

The thirty-second version. Expose Nothing.

Expose Nothing sits in your browser and watches the boxes you paste into. Before anything reaches ChatGPT, Claude, Gemini or any other web tool, it flags the personal data it can see and offers to strip it. No account, no server, no configuration — the scan happens on your machine and the page never leaves it.

🔒 chat.openai.com
Expose Nothing
Summarize this engagement letter for me:
Client [CLIENT]
Contact [EMAIL]
IBAN [IBAN]
Matter [CASE-ID]
The extension, mid-catch — right before the paste goes out

What it catches

Names & contacts

Person names, email addresses and phone numbers in any of the text you are about to send.

Financial identifiers

IBANs, account numbers and card numbers — the fields that turn a careless paste into a reportable incident.

Case & client references

Matter numbers, file references, VAT and registration IDs in the formats your sector actually uses.

Your own sensitive terms

Add the words that are only sensitive to you — a client name, a codename, an unreleased product.

Running in under a minute

Step 01

Install from the Chrome Web Store

One click, no account, no card. It asks for the permissions it needs to read the page you are on and nothing beyond that.

Step 02

Open any AI tool

ChatGPT, Claude, Gemini, Copilot, or the LLM box inside a tool you already pay for. The extension does not care which.

Step 03

Paste as usual

Identifiers light up before you send. Strip them with one click, or keep the ones that are genuinely fine to share.

Add to Chrome — free

Free, and free of the usual catch: no account, no telemetry, no paid tier waiting behind it.

The platform

Detect. Anonymize. Answer.

Step 01

Detect

Every prompt is scanned for personal and sensitive data — names, emails, IDs, card numbers, plus custom terms you define.

Step 02

Anonymize

Each identifier becomes a safe placeholder before the request leaves your infrastructure. The mapping never leaves your side.

Step 03

Answer

The model answers on safe tokens; anonLLM restores the real values in the reply. You get the answer. The model got nothing.

No single model ever sees who, why, or what.

Your document is stripped, then split. Each provider gets a different anonymized fragment — none can tell what it's about or whose it is.

 
anonLLMstrip · split · route
↳ split across providers ↴
OpenAI
who? · no idea
Claude
why? · no idea
Gemini
what? · no idea

The documents, calls and calendars clients trust you with. Anonymized.

Law, accounting, healthcare, finance — the work is full of names and numbers you can't leak. anonLLM sits under all of it.

🔒 anonllm.app
Engagement letter · AI summary
Matter [CASE-ID] — retainer agreement
Client [CLIENT]
Contact email [EMAIL]
VAT ID [VAT-ID]
IBAN [IBAN]
Retainer € 12,500
5 identifiers stripped · mapping stayed on your machine
Document review — contracts, invoices, filings, case files
[CONTACT]
call · summarized privately
Discussed matter [CASE-ID]; caller sends [DOC] by Friday. No names left the device.
Calls & contacts — names never reach the model
Calendar — this week
09:00 Intake — [CLIENT]
11:30 Review [CASE-ID] filing
15:00 Call [CONTACT] re invoice
Calendar — client names redacted before any AI reads your week

Built for teams who can't leak. Infrastructure, not another app.

F·01

PII redaction

Names, emails, phone numbers, IDs and card data stripped in real time — plus your own sensitive terms.

F·02

Any model

Works in front of OpenAI, Anthropic, or your own private LLM. Switch models without switching your privacy layer.

F·03

EU-hosted / on-prem

Run it in the EU or inside your own network. Data residency you can point to in an audit.

F·04

End-to-end encrypted

Encrypted so even we can't read it. Only you hold the keys — anonLLM and d2b never see your prompts or your data.

F·05

Never trains the model

Public models never see raw data — so there's nothing to train on, log, or hand over.

F·06

GDPR / DSGVO ready

Designed around EU data-protection rules from the first line, not bolted on later.

Which one do you need?

Most people start with the extension because it costs nothing and works today. You outgrow it the moment the requirement stops being personal and starts being organisational.

Expose Nothing anonLLM
Price Free Private beta — talk to us
Where it runs In your browser EU-hosted or inside your own network
Covers What you paste into web tools Every prompt your systems send, including API traffic
Who it is for One person, right now A team that has to prove it
Audit trail Yes — what was stripped, when, by whom
Setup Under a minute An afternoon, with us

Give your team the AI. Keep the data.

⚖ Law firms∑ Accounting & tax♁ Healthcare€ Finance & insurance⬡ Anyone under GDPR
✓ You're on the list — we'll be in touch when your spot opens.
Private beta · one email when doors open, nothing else.
Not ready to talk? Take the free one instead — it works today. Add to Chrome — free ↗