Classify the risk: what data is at stake?
Workforce AI Security classifies every prompt in real time. Your job: read each prompt and tag the sensitive data types. Click the prompt to reveal the answer.
0
Correctly classified
0 / 10
Prompts reviewed
Data classification categories
| Category | Examples | Typical Action |
|---|---|---|
| PII | Names, SSNs, DOBs, addresses, phone numbers | Redact or Prevent |
| Credentials | API keys, passwords, tokens, certificates | Prevent (always) |
| Source Code | Proprietary code, internal libraries, config files | Prevent or Ask |
| Financial | Revenue, forecasts, M&A data, salaries | Prevent or Redact |
| Medical/PHI | Patient records, diagnoses, treatment plans | Prevent (HIPAA) |
| Confidential | Unreleased plans, partnerships, internal strategy | Ask or Detect |
| Safe | General queries, public info, learning questions | Allow |
Loading...
Loading...
Think Deeper
Try this:
A prompt contains a customer name AND an AWS key. Which classification matters more? Why?
The credential is more critical — a leaked AWS key provides direct access to infrastructure and can be exploited within minutes by automated scanners. PII exposure is serious but typically requires social engineering to weaponise. Real systems use highest-severity-wins for multi-label classification.
Cybersecurity tie-in: This classification task is exactly what Workforce AI Security does
automatically on every prompt — using NLP models trained on millions of examples.
The same supervised learning approach from Stage 1, running inline at millisecond latency.