Card: for education only, no claim to subject-matter expertise; links to BrewSLM and the build write-up

README.md

@@ -14,7 +14,16 @@ language:
 # Sensitive-data tagger (PCI / GDPR / HIPAA + custom) span tagger
 
 A token-classification model that finds spans of text and labels each with a type, fine-tuned from `microsoft/deberta-v3-base`.
-Trained with BrewSLM (project “Sensitive-data tagger (PCI / GDPR / HIPAA + custom)”, run #143).
+
+> **For education only.** This model is a demonstration of a training method. It is not a privacy, security or
+> compliance expert, and it makes no claim to subject-matter expertise. It is not legal or compliance advice. It can
+> miss sensitive data and can tag text that isn't sensitive. Don't rely on it as a safeguard for real personal data:
+> consult a qualified privacy or compliance professional.
+
+**Trained with [BrewSLM](https://brewslm.com)** (project “Sensitive-data tagger (PCI / GDPR / HIPAA + custom)”, run #143),
+a platform for building small language models from your own data. How this model was built, step by step:
+[*Build a sensitive-data detector from zero, with no labelled data*](https://brewslm.com/blog-sensitive-data-detector-from-zero.html),
+and more on the [BrewSLM blog](https://brewslm.com/blog.html). Non-commercial showcase; no warranty.
 
 ## Model
 
@@ -147,6 +156,8 @@ def find_spans(text):
 
 ## Limitations
 
+- Educational demonstration only, with no claim to privacy, security or compliance expertise. Not legal or compliance
+  advice.
 - Scores come from small held-out sets — 44 texts (answer key), 44 texts (test examples) — so expect a few points of noise either way.
 - Types with fewer than 10 reference spans in an evaluation (`address`, `api_key`, `bank_account`, `cvv`, `date_of_birth`, `employee_id`, `health_plan_id`, `iban`, `ip_address`, `medical_record_number`, `passport`, `routing_number`, `ticket_id`, `username`) have unreliable per-type scores.
 - The training data is synthetic or model-labelled, not human-annotated; real documents with other layouts, formats or languages may score lower.