Blog
PII redaction, in practice
Guides, war stories, and honest comparisons on keeping personal data out of LLM providers. Written by the people building Sether.
PII redaction for LLM applications: the complete guide (2026)
How to keep names, emails, cards, and secrets out of OpenAI, Anthropic, and other LLM providers: detection methods, reversible tokenization, streaming pitfalls, and tool comparison.
- ·9 min·typescript
How to redact PII before sending prompts to OpenAI (TypeScript)
A hands-on TypeScript tutorial: detect and redact emails, phones, cards, and API keys before they reach OpenAI, then restore them in the reply. Streaming included.
- ·8 min·python
How to redact PII from LLM prompts in Python
A Python tutorial: strip emails, phones, cards, and secrets from prompts before OpenAI or Anthropic sees them, then restore them in the reply. FastAPI and Flask included.
- ·10 min·gdpr
Is it GDPR-compliant to send customer data to OpenAI?
It can be, if you sign the DPA, list OpenAI as a sub-processor, handle the US transfer, and apply technical measures like redaction. Here is the honest checklist.
- ·11 min·presidio
Microsoft Presidio alternatives for Node.js and TypeScript (2026)
An honest comparison of Presidio, LLM Guard, Sether, and DIY regex for PII redaction in Node/TypeScript apps, including when Presidio is still the right choice.
- ·8 min·pii-redaction
Reversible PII redaction: how redact-then-restore works
Why one-way masking breaks chat UX, how stable tokens and a local vault make redaction reversible, and when reversibility is the wrong choice.
- ·9 min·shadow-ai
How to stop pasting sensitive data into ChatGPT (without banning it)
Bans push AI use into the shadows. A browser-side guard that scrubs PII before the prompt is sent works with people instead of against them. How it works and its limits.
- ·9 min·streaming
The chunk-boundary bug: why streaming PII redaction is hard
How a property-based test caught emails slipping through Sether's streaming redactor when values split across chunks, and the hold-back window that fixed it.
- ·10 min·ner
Regex vs NER for PII detection: when patterns aren't enough
Where validated patterns win, where label anchoring helps, and where only NER catches PII. Trade-offs, local ONNX inference, and a practical recommendation ladder.
- ·9 min·supply-chain
How I took my npm package's supply-chain score from 78 to zero alerts
Socket.dev scored Sether 78 from 7 devDependency alerts. How I fixed all of them, dropped to 1 runtime dependency, and why a security tool must hold a higher bar.