AI Privacy for Individuals

Why personal AI use leaks the most

Personal ChatGPT use is where the most intimate data ends up in someone else's logs. A medical question with a real symptom. A salary negotiation pasted into Claude for advice. A draft email mentioning a family member by name. A cover letter with a home address. None of that should sit on an AI vendor's servers indefinitely – but by default, it does.

SOWA Privacy sits between you and the chat interface. Your prompt is scanned in the browser, personal data is replaced with placeholders, and only the sanitised version reaches the AI. The AI's answer is mapped back to the original terms locally, so you still get a useful, specific reply – without the AI vendor ever seeing the raw text.

What SOWA detects

Three layers run locally in your browser, before a single character is sent: a regex layer for structured identifiers, an optional multilingual NER layer for names and places, and a user-managed blacklist for the terms specific to your life.

Personal identifiers

Structured IDs caught by the regex layer.

email addressphone numberhome addressdate of birthtax IDpassport no.driving licence no.IBAN

Family & relationships

Names and roles caught contextually by the NER layer.

your namepartnerchildrenparentssiblingslandlordemployer

Finances

The numbers behind every money question.

salarybank accountcredit cardloanmortgagerentnet worth

Sensitive context

Health, beliefs, and other Art. 9 categories.

medical conditionsmedicationsreligionpolitical viewssexual orientationtherapist notes

Privacy is the default, not a setting

Nothing leaves your device

The detection engine runs entirely in the browser. No analytics, no telemetry, no "anonymised" usage data – there's no server to send it to. The optional NER model is downloaded once from HuggingFace and then runs offline forever.

You stay in control

Add the terms you care about to the Blacklist – your pet's name, your address, an old phone number. Add things you don't want flagged (your own GitHub handle, a public alias) to the Whitelist. Settings live on your device and sync with your browser profile, not with us.

Free, forever, for personal use

The full detection engine is in the free tier. Paid plans exist for organisations that want centralised policy management – individuals get the same protection at no cost.

Set it up in two minutes

  • Install the Chrome extension from the Web Store (or load it unpacked – the source is on GitHub).
  • Open ChatGPT, Claude, Gemini, or any other AI tab – the extension activates automatically.
  • Type a prompt with personal data. Watch the names, addresses, and numbers swap to placeholders before submit.
  • (Optional) Open Settings → Detection → Custom rules & lists to add your own terms.

SOWA Privacy is a privacy tool, not a content filter. It anonymises before the prompt leaves your browser; the answer you get back is still as useful as the question you asked. The only thing missing from the AI vendor's logs is which person the conversation was about.