Dev Tools · 1h ago
Corpus-scrub 0.1.0: Open-source tool redacts PII and secrets from training data
Corpus-scrub is a local-first CLI that detects and redacts personal information and secrets in LLM training corpora before training. It uses Presidio NER for names and custom regex for emails, phone numbers, IBANs, and credit cards, plus a ported gitleaks ruleset for secrets. The tool outputs masked, hashed, or dropped data with an audit report, and is licensed under AGPL-3.0.
Meridian48 take
A practical, privacy-focused tool that addresses a real compliance need under GDPR and the EU AI Act, though its recall guarantees are limited to certain PII types.
Read the full reporting
corpus-scrub 0.1.0: detecta y redacta PII y secretos en corpus de entrenamiento antes del entrenamiento LLM →
DEV Community
pii-redactionllm-training-data